Lessons Not Learned

The Department of Defense’s (DoD’s) Space Development Agency (SDA) hosted an Industry Day last week and presented its plans for deploying 20 satellites in two near-polar low Earth orbits (LEO) by 2022. The proposed satellites are part of the SDA’s demonstration plan for “Transport Tranche 0” and intended for communications (the DoD calls this a “transport layer”), with payloads dedicated to the DoD’s integrated broadcast service (IBS--not irritable bowel syndrome) and Link 16 systems. If those work well, then the SDA would orbit hundreds of similar satellites and then transfer control of them to the U.S. Space Force (SF).

Of note, the satellites would be using Ka-band frequencies for communications with the ground--but, they might also use optical inter-satellite links (OISL--laser communications) to pass information between the satellites. If those OISL’s are too finicky, then maybe V-band radio crosslinks would be integrated instead.

The slide below comes from one of the 86 slides the SDA provided:

An eye-chart showing how the SDA sees the world. Below is a chart that is more relevant (and more clear) to this article’s analysis:

Gosh--plans to implement a brand new type of large military communications constellation that would require a large ground infrastructure using still expensive and exotic technology such as laser communications crosslinks dedicated to one, maybe two, military communications systems.

Some of this sounds familiar.

Actually--all of it sounds familiar--and very unrealistic. Especially the constellation’s capabilities and schedule. Going back to one of many interesting conclusions from a 2006 Government Accountability Office report:

For the most part, this has not been caused by poor cost estimating itself, but rather the tendency to start programs before knowing whether requirements can be achieved within available resources.

That report was referring to military space programs like SBIRS, AEHF, etc. But it seems as if it could be applied to this SDA space Transport architecture. The requirements the SDA displayed in this slideshow don’t appear to be achievable with available resources.

Schedule "Risk"

Tackling the most unlikely prospect first: the schedule. The SDA would like to have 20 satellites operating by 2022. This schedule includes actual satellite bus and payload designs as well as selecting the ground architecture that makes sense for the constellation. Then it means selecting the contractor who has the best shot at creating the satellite and ground infrastructure.

None of the current U.S. primes--Boeing, Lockheed Martin, or Northrop Grumman--have demonstrated the ability to build one sophisticated military communications satellite in less than three years, much less a constellation of 20. To suggest they could then manufacture and deploy “[h]undreds of Transport Layer space vehicles” (the initial estimate was for 250) in less than four years from today (with potentially other mission-unique payloads) is ludicrous. Okay, maybe it’s not impossible, but based on other factors, including history, this proposed scenario is very, very unlikely.

This SDA schedule would require legacy satellite manufacturers to revamp factory floors and take hard looks at their processes to accommodate its schedule. That would cost those companies money, which they would charge back, somehow, to the DoD. These changes are not inexpensive and they aren’t something that can be done quickly through typical defense acquisitions processes. Why do I focus on legacy manufacturers--because the history of DoD acquisitions for these kinds of space projects tend to favor them.

Looking at a few of the latest commercial LEO broadband contenders with relevant experience and technologies may provide some common points for comparison. OneWeb (now bankrupt) constructed a factory in Florida to build two OneWeb satellites a day. It cost OneWeb about $85 million dollars to build a new factory for its satellites. It also took the company over two years from factory ground-breaking to opening its doors. Only half of its ground stations were in place by the time the company pulled the plug in March 2020.

To reiterate: It took OneWeb two years to build a factory. The company was racing against time and some very aggressive competitors. It’s probably realistic to assume OneWeb was pushing its contractors hard to get things done quickly. And it still went bankrupt. This very recent OneWeb history seems to indicate that the SDA’s schedule is...optimistic. Unrealistically optimistic.

Future Communications Architecture for the Space Force?

The SDA’s program schedule is impacted by more than building/revamping a satellite factory. There’s also technology. OneWeb relied on a global supply chain across four continents. U.S. laws will make it unlikely for such an expansive supply chain for DoD satellites. This may mean the supply chains for the SDA architecture are given a monopoly, which increases costs. This can impact certain satellite payload characteristics for the SDA that a company like OneWeb prized and took for granted, such as data transmission speeds and low latency.

Using OneWeb’s satellites as an example for some of the components and constellation design the SDA seems to be demanding for its satellites. Based on what is being presented, it appears that OneWeb’s satellites are more sophisticated in payload design than the SDA’s satellites. They are quicker than SDA’s projected speeds-- 9 Gbps for a single OneWeb satellite versus 250Mbps low latency transfer for an SDA satellite. SDA’s low speed selection might be because terrestrial U.S. military equipment (and maybe a few military satellites) aren’t designed for higher bandwidths. It may be because the SDA is limited to U.S. suppliers for its satellites.

To the point of low latency...traditional communications satellites in geosynchronous orbit have latencies of 600+ milliseconds. LEO satellites, because they are lower, have a natural latency advantage when comparing them to GEO satellites--about ~20-50 ms. SpaceX’s first batches of Starlink satellites are functioning at about half the proposed SDA altitude with latencies at sub-20 ms. This difference indicates that the 1,000 km altitude proposed for the SDA is a compromise, as the higher altitude means a higher latency. Why wouldn’t the SDA go for sub-20 ms latency to support U.S. troops? Probably because that altitude would require thousands of satellites.

OneWeb did not pursue crosslinks for its constellation. There are several good reasons for OneWeb not to have done so, but probably the biggest reason was to keep costs at ~$1 million per satellite--especially when considering OISLs for satellites. OISLs are still somewhat rare for spacecraft to spacecraft communications and will likely cost the DoD a pretty penny. Will OISLs provide enough of a glimpse of the benefits of their use in a constellation of 20 satellites in near-polar orbit? How many optical modules will be required per satellite? Four? Are there any U.S. OISL manufacturers capable of producing 80 optical modules in less than three years? Or a thousand in less than four?

A big potential difference between the SDA’s satellites and OneWeb’s will probably be the per-satellite costs. It is doubtful the SDA’s satellites will be as low as a million dollars each--especially if it chooses a legacy manufacturer. But it seems that if the SDA continues down this path, its satellites SHOULD BE closer to OneWeb’s $1 million per satellite price than $100 million. But history gives reason to doubt that will happen.

Another option would be to wait and see what happens with SpaceX’s Starlink constellation. Maybe experiment with it. It would be less expensive and quicker to use than “Transport Tranche 0.” It would give the agency experience with LEO networks. It might be able to use SpaceX for its satellite manufacturing down the road, should it still decide there is a need for an SDA constellation. SpaceX can manufacture seven satellites per day.

Instead what the SDA appears to require for its satellites are technologies that are far from fast. The satellites will not be orbited in a position that provides the lowest latency possible. Looking at the DoD's past performance with acquiring these systems, as well as what appears to be the current state of satellite manufacturing in the U.S., it will also be late. This is all very odd, considering these satellites should be providing an edge to U.S. troops and other DoD/national assets.

These satellites will include technologies, such as laser crosslinks (which are fast), perhaps solid-state antennas (my conjecture), that are very expensive and may not be mature enough to have fast production lines. The finished constellation of 250 satellites will not be “future-proof.” It might not even provide adequate coverage for global military operations. And this is only when compared with OneWeb. SpaceX’s Starlink satellites appear to be more capable, faster, and lower latency for about ⅓ the cost of an OneWeb satellite.

Maybe that’s a jump too far into the future for the SDA.