Starlink on trains

“It’s often difficult to identify. I know the start point of when an idea becomes a conversation becomes a project, but can you just talk me through exactly what the problem that needed solving was?”

It’s very difficult to actually make a phone call on a train. It’s very difficult for people who are travelling on trains to actually work consistently when they need to be able to get online and there’ve been many attempts over the last 20 years to actually provide connectivity onto trains, starting from the assumption that you could use your own mobile phone, and alternatively that the train operators could provide Wi-Fi and that has begat the problem.

How do we actually provide connectivity to the train? Or to the people?  The precise conversation before why we wanted to put star link onto a train really started in 2023 and that’s because Starlink testing had been completed in terms of providing connections demonstrating that they could put a Starlink antenna strapped onto the roof of a of a pickup, or a four-by -our and use it as a form of mobile terminal and that they could actually achieve significant throughput on those terminals?

Well, wouldn’t it be a good idea if we strapped one of these to an aircraft?  So they strapped one of these terminals to an aircraft, which necessitates a significant amount of engineering development to make it suitable to actually fly on a plane?

Then in the summer of 2022, a value-added reseller Claris Networks were asked if they could actually assist in making the transition to taking the terminal onto a train.

“So would it be fair to say that the issue was more about reliability than speed?”

Yes, it would be fair to say that’s the case, but the issue has oscillated over the last 20 years with iterations of 2G, 3G, 4G and 5G.

It’s also oscillated because of coverage, capacity and expectation. Expectations were incredibly low 20 years ago and the sales pitch was equally low. Expectations are now significantly higher about what you can do, since people now expect to be able to stream something whilst they’re sat on the train. In many cases they also expect to be able to. undertake a live Teams or equivalent while they’re on the train, or worse still, they want to actually make a live phone call!

And it’s the live phone call that’s consistently demonstrated that the expectations and capability to deliver really haven’t meshed very well at any point until the idea of having a solution that’s available with widespread, almost ubiquitous coverage, that could be provided.

“So you talk about almost ubiquitous coverage – can you put a figure on that in terms of the terrain that the UK rail network travels through? What does Starlink offer in terms of that ubiquity  – tunnels notwithstanding?

“99% and the remaining 1% is accounted for by tunnels, and some stations which, like Birmingham New Street, is essentially built in a concrete bunker.

So, it would be fair to say that it is ubiquitous coverage unless a man-made obstacle gets in the way.

“That’s a really helpful statement of both the problem and the solution that Starlink offered – How did BAS get involved?”

Clarus, which was relatively small at the time, were faced with a problem. That being that they were asked by SpaceX to take the aviation product and put it on a train, which meant it had to be certified. Clarus knew that they didn’t know how to certify it, but they felt they knew somebody who did!. I was asked if I could pick up that problem for them and the answer was yes, because we felt that we could actually support Clarus to take this product and put it into the testing programme as well as helping SpaceX to understand what modifications they would need to make to complete the testing successfully.

“So in that sense, you were a perfect partner for Clarus, in that you operate within the telecom sphere and have decades worth of experience of that, plus knowledge of work in the rail industry in a telecoms context, plus knowledge of the regulatory environment within the UK and particularly within the UK rail network that actually had to be navigated – I’m assuming a navigation that is not always easy?”

We were in the right place at the right time and had a significant proportion of the knowledge that was required. Plus, we were prepared to say “yes, we can do it!”. It was really picking it up and running with it and recognising that there would be questions that we couldn’t answer – the big question being whether SpaceX could actually answer the questions that we couldn’t.

I think my previous experience is that I’d worked on the railway a lot as a supplier and so I had taken products through Network Rail product approval, which is slightly different to a product approval on the on the actual rolling stock itself, but largely the same principle where you had to “shake it and bake it” and really just try to break it. So, I’d got a reasonable amount of experience of doing that from previous roles.”

“And what does Clarus bring to the party?”

Mainly the fact that they’re a value-added reseller for Starlink, delivering a service rap for the Starlink antenna and provide a commercial element in terms of data pools and all of the other items that comprise the cost to a train operator to have Starlink on their train. They also provide the interface to the firms that can install Starlink on the train and then operate it as part of the onboard system. The Starlink antenna is one part of a significantly sized multi part system, which is how do you connect the train to the outside world and having connected the train, how do you distribute that connection throughout the train and provide the connection to systems and passengers.

“So ostensibly dealing with Clarus within the UK, is as if they are Space X’s representative on Earth?”

Yes, they are one of a very small number of approved value-added resellers They also have the engineering expertise to be able to demonstrate the performance of the equipment before the tests and then demonstrate the performance of the equipment after the tests that allowed us to prove that the equipment could withstand the tests required.

“Presumably, within that iterative process, they also have the engineering experience to amend the equipment if required as a result of the tests?”

They knew there would have to be adaptations, yes. We were working as an embedded part of Clarus to actually identify where the problems were and encourage SpaceX to make the amendments to the aviation product to make it a certifiable locomotive product.

Basically, we led it directly with SpaceX and we’ve had a weekly call with the SpaceX team out in California and Seattle for the last 12 months.

“OK. So can we move on to what actually is required to get a piece of kit like this certified to, in layman terms, stick on top of a train? What are the regulatory considerations and what did you have to do to nursemaid the kit through that process on behalf of the partnership?”

So there is a standard, a British standard, called 50155, and that basically tells you what tests you have to do. Essentially you have to check test its DC power, test it in extremes of temperature, test it in extremes of salt water, examine its electromagnetic compatibility and how it copes with very, very rapid changes in temperature to emulate going, let’s say, from a cold environment into a tunnel very quickly where it might be 10° warmer and then straight away back out to a cold environment again.

You also have to simulate, say, coastal areas by spraying it with salt mist solution for quite a number of days and, critically, you have to then check its performance to assess the amount of smoke and the amount of toxic gases it gives off when burning.  That then itself will dictate whether it can be used in tunnels of up to a mile or up to a mile, or up to five miles, or beyond 5 miles. So, it’s a really quite a lengthy list of tests that you have to follow.

We were using a test house in the UK, so we had to produce the overall test plan. Based on the test requirements for the overall standard, we then produced an individual test listing the elements we needed the test house to conduct. I think it took about six months to go through all of the test cases and then at the end you then have to go to an independent organisation to attest to the fact that all the tests that you’ve specified and carried out are in line with the actual specification.

“And how did the how did it fare within those tests? Obviously not wishing to plot spoiler it passed, but did it pass everything first time, or was there quite a lot of iteration?”

So to be fair, the actual product itself performed really, really well.

One of the other questions we had to answer, because this will go under the overhead lines, which are at 25,000 volts, was “what happens if one of those overhead lines falls onto the onto the actual antenna itself? Will that 25,000 volts be transmitted down into the train unit itself? Will it set fire to the unit? Will it destroy it? And so we had to replicate, as far as we could, 25,000 volts falling on to the antenna.”

“Did you did manage to do a test for that?”

Yes, it was a simulated test and they literally had to put a node on the actual platform itself, and subject it for so many milliseconds to so many 10s of thousands of volts!

“Talk me through the key milestones that led up to where we are today?”

The key milestones were identifying the tests required, producing a test specification and identifying test locations to perform for quite a period of time, then carrying out those tests and then appointing an attestation organisation to then sign off those tests. Also get the final Test results in.

There was also one precursor milestone, which then continued as a stream of work in parallel, and that was “is there demand for this and, if there is demand, how big is it and what level of sales might SpaceX and their resellers, expect to deliver. The biggest challenge here is to actually qualify and then quantify this by understanding how many terminals you’d expect to put on to an individual train, for instance. Because this in turn is determined by what you’re going to do with it and what services you’re going to deliver to the train and or crew and passengers on the train.

“I recall from previous conversations that this involved an analysis of the European network, the volume of traffic, the number of trains operating on the overall network, therefore the potential size of the market – all of that kind of good stuff?”

Yes, absolutely. So, there was the market analysis plus the fact that there was already a candidate customer, ScotRail. ScotRail had already asked to be able to provide connectivity from Inverness up to Thurso on the Far North line. So, there was a known demand but the issue was how extensive was that demand and how extensive will it be going into the future. I’m pleased to say that the indications now are significantly stronger than they were 12 months ago, when we were estimating interest. In fact, there’s a lot more interest and a lot more understanding that this can deliver an answer without having to expend large sums of money on trackside infrastructure.  Also, in cases where the train operating company doesn’t have any influence over stuff that might go in the trackside, not just in this country but elsewhere in the world, it provides it provides the basis for resilience solution, particularly if you take a take a tile from a different satellite constellation to provide what is potentially a rock-solid connection to a train.

“The next question is therefore, where are we now? What stage is the project at?”

So the device is now ready and certified. And is due to be installed on the 1st trains for ScotRail in January and then we will have live results from those trains once it’s been installed and they’re back in traffic.

“OK, so a couple of questions emanating from that. I think I think I can guess the answer to the first one. But why ScotRail and why that particular route?”

So why ScotRail? For a solution on the Far North line, which is the route that goes from Inverness to Thurso. There’s no mobile network coverage, there’s no GSMR, which is the GSM for railways solution, alongside the track on that on that route and they use a curious signalling system, radio electronic token block, which actually means that for significant periods of time they don’t know where the train is, other than the fact they know it’s between point X and Y.  They don’t know where it is and there’s no way of communicating with the driver.  Clearly there’s also no communication methods for the for the passengers on the train. So, a decision was taken to come up with a solution that would work to provide a connection to the train for the safety and operational support of those trains and coincidentally provide some service for passengers.

“It sounds like you’re taking an awful lot of boxes there. Is there a balance between the health and safety elements, the network operation and the retail offering, if you like, to passengers in terms of priority or are they all relatively equal?”

From the train operators perspective, the bigger issue is about the safety of the train – It’s the operational requirements that are making the idea of putting Starlink on a train more attractive to the train operators, rather than relying on traditional methods, but once you’ve got that capacity, connecting to the train, you’ve then got the ability to also support the commercial benefits of providing a connection for passengers.. But, from a train operator perspective, it’s the operational side that’s driving the value of connecting their trains.

“So that’s ScotRail.  You talked earlier about your role in analysing the wider market need -where does that go, in terms of interest that may have already been expressed by other operators? How does this then get implemented and expanded?”

So there are several companies that want to do a proof of concept in the UK, in Europe, in North America, South America and Australia.  There are expressions of interest from all over the world to do some form of proof of concept, and they’re already a significant number of train operators at home and internationally, who are working up proposals to actually install and deploy the solution.

It’s also worth mentioning that the standard to which this has been tested is the standard across 34 countries in Europe.

“When we talk about the ScotRail example, obviously I can see all of the important attendant safety aspects, train operation aspects there. But how do you work out the provisioning for the passenger facing part of the network?”

That’s a good question because the size and therefore passenger carrying capacity of trains varies enormously from a single carriage train carrying up to 40 passengers through to the monsters that run through London in the in the commuter role of carrying between 1100 and 1600 people. So, the bit that’s common is if you’re providing simple Internet access, you’re going to have Wi-Fi that’s accessible to customers and that’ll be on the back of a local area network running through the train. The difference will be what is the available capacity that they can then tap into and the more practical way of saying what it’s likely to be is to identify how many antennas you might realistically put on a train roof and the cost, and then it will be a determination of what passengers will get. Rather than saying that all passengers must get, for the sake of argument, 100 megabits per second, the practical reality is it’ll be dividing what’s available and understanding whether that’s going to deliver. something that’s a significant improvement to customers.

“So, in simple terms, it is the case that 2 tiles would provide twice the capacity and four tiles would provide four times the capacity, and then it becomes a commercial decision for the train operator.  Yes, to do the maths, analyse their usual occupancy per train per carriage and work out how many tiles and work out what level of service they want to provide, the extent to which they want that to be a value-added proposition that they can heavily market to their passengers and therefore what level of investment they would need to make. To provide that that service in terms of how many tiles, one per carriage, or 2 per carriage, if they wanted to go large.