The Galileo Eccentric Satellite Surprise

UPDATE: Back in December 2020 when this article was first posted, I noted that some receivers might not be able to deal with the advent of production signals from the eccentric satellites. This has indeed proved to be a problem for U-blox F9 series receivers in a specific configuration. The application note links to this article (thanks!) and also (correctly) states that the signals from the eccentric satellites do not in fact comply with the Galileo Open Service ICD. I don’t think it is fair to blame U-blox for having problems with non-compliant data.

It appears the EU GNSS Agency agreed, and from the 18th of February has ceased transmitting production data from the eccentric satellites. Additional background is in this service note #05.

The rest of this post dates from the 7th of December 2020.

Last week, Galileo unexpectedly started transmitting production signals from two eccentric satellites. In this post I explain the background of the situation and the unique orbit of these satellites. I also cover why these new production signals may be good news, but that the communication by Galileo was somewhat atrocious, a situation made worse by denials that the communications were bad. Informally, I have now heard that it is acknowledged things were not good enough, and that improvements have been requested.

I start off with some history, then describe the ways in which E14 and E18 differ from normal Galileo satellites and I round off with the unexpected events of last week. If you already know the history and technical details, please feel free to skip to the last section ‘Last week’.

Percentage of receivers using the new Galileo satellites

Percentage of receivers using the new Galileo satellites

History

On the 22nd of August 2014, now more than six years ago, two new Galileo satellites were launched by Arianesat from Kourou. GSAT0201 and GSAT0202 were the first two “Full Operational Capability” satellites, and were launched together by a Soyuz-STB Fregat-MT rocket.

Sadly, an upper stage fuel line froze because it lost heat to a nearby helium line. On investigation, it was found that a quarter of Fregat upper stages produced so far had these lines close together/thermally bridged. Previous launch profiles may have provided enough heat to the fuel line to prevent freezing. More details can be found in these excellent two “Galileo GNSS” posts.

The satellites ended up in a very bad orbit, one that posed radiation dangers for the satellites. When satellites are launched, an intense period of “babysitting” (LEOP) is planned where all expertise is available to help ailing satellites. And not without reason, it is extremely common for a satellite to get into trouble immediately after launch. Expert help is then on hands to save spacecraft from dangerous radiation, too much / too little sunlight or even burnup.

With great heroics, GSAT0201 and GSAT0202 were saved and brought to a safe orbit. And with further clever work, the satellites were moved to an eccentric orbit that might be suitable for actual operational use for navigation and timing. Starting the 5th of August 2016, the satellites started broadcasting testing data as Galileo satellites E14 and E18. This was quite an achievement.

There are three unique aspects to the E14 and E18 orbits: the relativistic effects of the elliptic orbit, the precise maths required and finally the almanac:

Relativistic effects of the elliptic orbit

GPS, Galileo, BeiDou and GLONASS satellites have always orbited in very circular orbits. To get accurate GNSS performance, the orbit and clocks of satellites need to be modeled very carefully. This is a lot easier if the orbit is as near circular as possible.

Every satellite transmits an ephemeris which describes the orbit and clock for the next few hours. Even a perfectly circular orbit deviates a bit because of solar pressure, uneven gravity fields and other reasons. In addition, clocks do drift, but mostly they do so at a known rate. So the ephemeris includes drift parameters as well.

One effect is not reflected in the ephemeris - the influence of general relativity. Time runs slower in a gravitational field. If an orbit is ever so slightly eccentric, the satellite spends half of the orbit in a slightly stronger gravitational field than in the other half of the orbit. Or put another way, during a full orbit, the clock will first run slightly ‘ahead’ of where it needs to be and in the second half it will drop behind by the exact same amount. Averaged out over a whole orbit, the clock speed is not impacted by the eccentricity. But during the orbit, it is.

Even for the near perfectly circular orbits of regular Galileo satellites, this effect is one or more nanoseconds, or, dozens of centimeters of ranging accuracy.

For the eccentric satellites E14 and E18 this effect is over 200 meters! Receivers have to carefully apply this correction. If a receiver was not doing this (well) for the circular orbits, this was not too bad. But messing this up for an eccentric orbit is catastrophic.

In addition, the eccentric orbits are not described very well by the broadcast ephemeris. Where a regular circular Galileo orbit is described correctly for many hours by a single ephemeris, the eccentric descriptions quickly lose accuracy, even after only a single hour.

Incidentally, the launch of multiple very precise atomic clocks in these eccentric orbits did enable some very good science. Using the data from E14 and E18 it was possible to determine the accuracy of one prediction of Einstein’s theory of General Relativity by a factor of four beyond what has been achieved previously.

Kepler

The ephemeris broadcast by satellites describes in detail where we can find the satellite for the next few hours. Interestingly, once orbits are not precisely circular, describing the true anomaly (“where you are on the ellipse”) is not a trivial calculation.

Galileo provides several suggestions on this front. The Galileo ISD states “Kepler’s Equation for Eccentric Anomaly E (may be solved by iteration)”, and then provides the following somewhat odd formula to get to the true anomaly \(\nu\):

\[ \nu = \tan^{-1}\frac{\sqrt{1-e^2}\sin{E} / (1 - e \cos{E})}{(\cos{E} - e)/(1- e \cos{E})} \]

In addition, a page on the Galileo Service Centre website offers another way to calculate the true anomaly \(\nu\):

\[ \nu = M + e\ 2 \sin(M) + e^2\frac{5}{4}\sin(2M) - e^3\big(\frac{1}{4}\sin(M)-\frac{13}{12}\sin(3M)\big) + \cdots \]

It notes this formula has been supplied by ESA.

If you now go use the first formula, you discover quickly that from time to time it provides answers with the wrong sign. The second formula however is always right, so you might end up using that one. But the dots at the end of that equation should be a hint that this is an approximation.

And turns out, it is an approximation that fails for the eccentric satellites, leading to 200 meter errors. We may hope receivers picked the correct formula, which is :

\[ \nu = atan2(\sqrt{1-e^2} \sin(E) , \cos{(E)} - e) \]

Almanac

Every Galileo satellite transmits details about its own orbit and clock. In addition, every satellite transmits the ‘almanac’, which tells receivers about all the other satellites. Originally, the almanac was required to even find the other satellites. These days, receivers can perform this feat without the almanac.

The almanac remains important however. To keep the size of the almanac manageable, it includes reduced resolution details of all the satellites. Specifically, since no one saw eccentric satellites coming, the almanac is unable to encode the orbit of E14 and E18. There simply are not enough bits in the field that stores the eccentricity.

So for this reason, E14 and E18 can not appear in the almanac. This has now been proven to be confusing to some receivers.

Last week

Recapping the previous sections: The eccentric satellites (E14 and E18) are quite different from the regular Galileo vehicles. Their ephemeris requires more precise math than regular ephemerides, and receivers might not be doing such precise math. This goes both for the orbit determination and the relativistic effect compensation. In addition, these satellites can not appear in the almanac, which could confuse receivers.

E14 and E18 spent four years broadcasting testing data, but the ’testing bit’ made sure receivers would not use the actual data from these eccentric satellites for positioning. Much Galileo monitoring infrastructure also did not measure E14 and E18.

Then, Monday the 30th of November, all of a sudden the the ’testing’ bit was removed from E14 and E18. This meant that billions of receivers (mostly phones) were now exposed to orbits they’d never used before.

This was not announced in any way. Many seasoned professionals were caught by surprise at this development. Normally, Galileo changes are communicated using NAGUs, ’notices to all Galileo users’. One can subscribe to these and receive them as email messages, and professional GNSS users do subscribe to this service. But no NAGU had gone out.

When we pointed this out, the EU GNSS Agency noted that they had published a Service Notice the Friday before on the page dedicated to such notices. But no notice had gone out to the subscribers of the NAGU update service. Nor had there been a news article or even a tweet.

I’m not aware how anyone would have noticed, unless they happened to have checked the Service Notices page by accident on Friday.

But even if you had, the notice.. did not in fact announce that something was going to happen!

The only thing in there was a description that these ‘auxiliary’ satellites might one day begin transmissions, and how one might use them.

The Monday after the Friday stealth launch of this non-notification, at 08:32 UTC the eccentric satellites dropped their ’testing bit’.

Then, around 14:00 UTC, finally, two NAGUs arrived in my mailbox describing what had happened.

Why?

Just why? This is the biggest change in Galileo operations since the 5 day disruption in July 2019. That too was communicated horribly. And now suddenly two eccentric satellites enter into service, without notification.

And to make matters worse, a Friday non-notification is then presented as justification. But that just is not right.

Then on Monday the satellites go live, and five hours later there is the first acknowledgment that something actually happened.

Meanwhile, GSA personnel explain on Twitter that notifications are not actually required to precede the event - as long as they appear within 72 hours afterwards.

Bad communications are bad enough. But to then tell people this is actually completely normal only makes matters worse.

The ongoing uplinking problem

As noted before, Galileo currently lacks sufficient uplinking capacity. This means that frequently it is not possible to send satellites up to date orbit and clock information (ephemerides) as frequently as we would want. Work is in progress to improve this situation, but the work is not finished.

Within a few hours after enabling production use of E18 on Monday, it had to already drop out of circulation again because no satellite dish was available to provide the frequent updates that the eccentric satellites require.

This was a very ‘on the nose’ demonstration why it may not have been wise to enable the eccentric satellites before the uplink network has reached nominal capacity.

The GNSS Marketplace

It should be noted that chipsets have a choice of satellites to use. The maths are easier if a chipset focuses purely on one constellation, say BeiDou. Doing multi-GNSS calculations is simply harder (and risky).

Chipset manufacturers ship policies with their chips. These can be changed on demand.

If a GNSS provider, like Galileo, is not communicating well, chipset manufacturers might change their thresholds and parameters to prefer other providers.

Many users have recently observed that their receivers prefer to flock to GPS.

At the very core, it will have to be decided if Galileo is an engineering or a political project. We can not divorce the two, but it should be realised the GNSS marketplace wants good and timely technical information. And not politically convenient announcements.

I would highly recommend that Galileo (EU GNSS Agency, ESA, European Commission) stops doing big things without announcing them thoroughly in advance.

This requires formal notices sent out as NAGUs. It requires service notifications that people actually hear about, and that are not stealth launched the Friday in advance. It also requires expanding communications into less formal channels - news items, blog posts, magazine articles and perhaps even tweets about upcoming developments.

As another suggestion, perhaps E14 and E18 could have been trialed in production for a day just to see how receivers would react.

I sincerely hope that Galileo will recognize that it needs to do a better job, and I specifically hope that they stop telling us that whatever they did was good because it was in accordance to their service definitions!

Informally, I have now heard that it is acknowledged communications to end users were not good enough, and that improvements have been requested.

To end on a happier note, E14 and E18 have now been in production for a week and the sky has not fallen. The Galileo uplinking capacity has also increased over the past few days, which should help keep the satellites updated better.

Clear improvement in performance with E14/E18

Clear improvement in performance with E14/E18

So the result was probably good. I wrote about the good news in this week’s Galmon.eu Galileo Performance Report.

EU Space Week

I wish the Galileo community a productive ‘EU Space Week’, which I am sure will feature news on this increase in uplink and satellite capacity. The relevant ‘Galileo Service Status’ session is on Tuesday, December 8th, 14:30-16:00 UTC. This includes a Q&A.

Registration for this event is free, so head over to euspaceweek.eu to make sure you can attend (remotely). I know I will!