Completely unintentional, yes of course ...
The Galileo roll out can't happen fast enough.
An error in the Global Positioning System (GPS) network apparently triggered by the decommissioning of a US satellite last week has had knock-on effect across a number of UK industries, it has emerged. Many industries are reliant on on GPS software for a swathe of critical applications such as financial trading and precision …
I am not clear what the article is actually talking about. Was the healthy bit set but ignored by ground segment users? Or did the problems occur between the onset of the failure and its recognition by the control segment?
The implications are vastly different.
Because it's a radio signal that is traversing the earth, so the recievers further away would be behind those closer to Anthorn.
For most things people do that wouldn't matter - but this was a 13 microsecond glitch causing issues...
That's 4 km
The UK is larger than that (citation needed)
Most of the critical stuff is probably static, so extra delays could be calculated - but this is a very high accuracy failure...
Could an equivalent be achieved by requiring all radio stations to broadcast their position and the current (accurate) time?
Seems a lot more simple than chucking satellites into space, although it wouldn't have the same coverage area, obviously.
I assume I'm missing something obvious though, just a thought.
"requiring all radio stations to broadcast their position and the current (accurate) time?"
only analog stations, not digital
and you'd have to get round the problem of multiple transmitters per station -e.g. BBC Radio Lancashire has four FM channels on different transmission sites
Hmm... indeed. One of those low-risk:high-impact failure modes that crop up in these discussions is a solar coronal mass ejection that takes out a significant number of orbiting electronic devices. Unless we believe that Galileo, GLONASS, etc. satellites are better hardened against radiation damage than GPS, they don't constitute an effective backup (for that scenario).
As far as I remember (haven't looked it up) we're currently on the downslope of the 11-year solar activity cycle: I'm sure we'll have reduced our dependency on orbital electronics by the time of the next maximum. [Insert unwarranted optimism icon of your choice].
This post has been deleted by its author
A 13 microsecond slew stops the prevision docking of oil tankers??? I would hope they used something more accurate than /just/ GPS for docking procedures, e.g. ship borne proximity sensors ... or even harbour 'pilots' and lots of eyes!
All systems can fail .. if you rely solely on /one/ system for anything important then you will get burned at some point.
"if you rely solely on /one/ system for anything important then you will get burned at some point."
.... And with an oil tanker, that burn has the potential to be serious.
Torn between Get My coat Icon and the one I eventually went for.
Tankers are pretty big. That is a lot of mass therefore when docking, the amount of thrust in one direction is critical that even a tiny amount of apparent movement can be too much. So yep, 13 ms when you are using that to determine position (radio travels a long way in 13 ms) means that you can't accurately determine the amount of thrust to bring the behemoth to a halt at the right point.
This is all supposition, btw. http://sploid.gizmodo.com/incredible-overhead-view-of-an-oil-tanker-perfectly-doc-1725572780
So yeah. Backup systems are key. I have no idea how accurate the other systems are compared to GPS. I am surprised, though, that some kind of, oh I dunno, laser ranging isn't employed against specified dedicated targets on a dock.
This post has been deleted by its author
What's 'precision docking of oil tankers' if it isn't navigation?!
Okay technically it's pilotage but it's pretty much the same thing.
As an aside I've seen it used to synchronise frequency hopping radios and the operator was completly unaware the numbers on the display had anything to do with the geographic position. Still it was handy to know his phone number for when there was an exercise fire on the bridge and we had to navigate from the upper deck.
Mort offered "radio travels a long way in 13 ms"
Even in the corrected 13µs (as opposed to 13ms), it's on the order of 13,000 feet. About a foot per ns.
It'd be a pretty poor navigation system that depends on absolute time, as opposed to relative. Typically, it's all relative timing.
it's on the order of 13,000 feet. About a foot per ns. It'd be a pretty poor navigation system that depends on absolute time, as opposed to relative.
It'd be a pretty poor navigation system that depends on 1/6th of a dead king's armspan as a unit.
But the metre is based on being one ten-millionth of the distance from the equator to the north pole of an utterly insignificant little blue green planet orbitting a small unregarded yellow sun, far out in the uncharted backwaters of the unfashionable end of the western spiral arm of the Galaxy. Not much better, really.
> A 13 microsecond slew stops the prevision docking of oil tankers ?
That is actually huge -- the speed of light is quite fast, so it correspond to about 4km so regardless of if you use the signal to calculate velocity (just a error of a few miles/h would make the docking impossible), or as a relative comparison to a second source it makes a big difference
In docking, the ship just needs to get close enough to the dock that it can be completed by tugs pushing it in place or longshoremen with ropes and winches pulling it into place. Beside, any tugger should be familiar enough with their port and their ship to be able to guide ships into place without anything more than basic instruments.
I suppose this is a symptom of skills going rusty because the computer does all the work for them, much like modern aircraft were if something goes wrong, better hope the pilot can read the manual in time...
"Exactly. There were big ships like oil tankers and aircraft carriers and whatnot decades before there was GPS. Makes you wonder how they managed? "
and thats why we had the Torrey Canyon, Amoco Cadiz, Exxon Valdez.................maybe they didn't manage
"how many super tankers ran aground whilst docking?". Yes indeed, referring to docking is a bit silly, the real danger is at problematic parts of the sea where no referent points are visible (and rocks are not visible) . Take, for instance, the Baltic where lots of tankers move to and from Russia.
Short non dramatic answer - yes oil tankers have different systems - usually a laser based one.
Sorry, I mean "OMG, all the tankers are going to explode!"
The issue more is that the reliance on a space borne radio signal seems woven in to many aspects of daily life and critical infrastructure as a convenience more than anything else. I remember being told in a previous life that the reason I needed to learn to use a map and compass was because, comparatively, GPS is like looking up into space for the light from a torch strapped to a satellite whilst its directly in front of the Sun. An oversimplification, but accurate enough.
Radios break, radios fail, radio can be jammed. The issue seems to be more an over-reliance on super accurate synchronized timing sources. The more accurate you need something to be, the more likely you'll get an undesirable result after all.
I hope that the designers of high reliability, mission critical GPS receivers use more than the mathematical minimum number of sats (four, I believe) to establish position.
My little handheld unit can usually get eight or nine good GPS sats, plus half a dozen GLONASS on the average day.
He said: "A more effective backup for GPS is desperately needed."
Anyone else thinking that misses the issue - a more effective replacement for GPS is desperately needed.
Companies place their dependence on systems that can have their accuracy manipulated by a foreign entity. Yes I know that the commercial GPS services are supposed to come with higher accuracy & reliability service levels - but that didn't help in this instance.
Serious question as I've not had to do this for a long time - what ever happened to taking the atomic clock signals as the prime source?
My navigation gear was always synched to WWV and we'd pay attention to the clock slew across the different systems. My being very proactive with the systems seemed to be appreciated by the navigator and quartermasters and I'd hear about even minor glitches long before they'd graduate to career enders. And it kept the Captain happy which is always a consideration.
Their sales should take off...
Amusingly, the CSAC has a pin to input the GPS 1pps signal (a pulse precisely aligned with the exact 'top' of each GPS second) to discipline the atomic clock. So any system design would likely still have a connection back to the GPS Time.
Life just got a lot more complicated for system designers.
This post has been deleted by its author
MEMS devices used commonly in smartphones used to be in that price range too, when they were produced in quantities of tens of thousands a year.
The advantage to including an atomic clock in a phone would be a massive improvement in position acquisition time as well as some improvement in accuracy (especially altitude) but you have a chicken and egg situation. The only chance for the price of a CSAC to drop like six axis MEMS did to allow their inclusion in cell phones would be if someone commits to ordering massive quantities. So probably Apple would have to think it is worth including to drive the price down enough for widespread adoption, as they did with flash and MEMS.
Apple bought an outdated fab in San Jose from Maxim last year, there was some speculation as to exactly why since it uses a very outdated process and would only be suitable for things like MEMS devices. I suppose there's a chance it could be used to produce a couple hundred million CSACs a year - to give Apple the benefit of this improvement in the iPhone without driving up volumes in the industry to where it is cheap and easy for everyone else to quickly match them.
This year they increased processing power by 70% (benchmarks run by anandtech and others proved their claims were quite justified) so that's not a "slightly increased" amount, but the 6 was already 'fast enough' so there are diminishing returns in utility of faster SoCs even if they were able to do that every year, which they obviously can't.
They'd prefer to add something unique that Samsung and other competitors can't easily duplicate, I'm just not sure "faster and more accurate GPS positioning" is going to move the needle. Perhaps combined with improved MEMS that allowed dead reckoning to a high degree of accuracy for when are out of range of GPS they could have a winner - and that fab they bought would allow them to gain economies of scale for their needs while similar capabilities would be too expensive for the competition unless they were willing to commit to ordering fifty million a year (and even then require a few years to catch up)
So I wouldn't be surprised to see something that's unique and not very easy for the Android world to duplicate in the 7S. Whether it is something that people care enough about having to influence their choice of what phone to buy (whether it means getting people to switch to iPhone, or stop them switching from iPhone) well that remains to be seen I guess.
I used to manage private Stratum 1 NTP servers. They did indeed use GPS as Stratum 0.
Proper NTP (as opposed to, say, Simple NTP) corrects for clock differences by accelerating or decelerating the local clock to gradually bring it back into harmony with the reference time.
I'm not sure that even El Reg's blind suicidal harbor pilot would notice a few ms drift over 12 hours.
I have an interesting story about NTP. Suppose you have redundant pairs of NTP servers, and hosts are configured to use both, in case one is offline. Now suppose that one of your NTP servers doesn't go offline, but just throws a wobbly and gets "stuck" at the wrong time. Many NTP clients (at least, those running the standard ISC code) will exit if the difference between local clock and the reference clock is too implausibly great. In this hypothetical case, there would be just a minor hiccup on your (redundant) NTP infrastructure with no downstream effects...until that threshold is reached, at which point thousands of systems would start to unrecoverably fail their NTP clients as they randomly hit the wobbly NTP server. Sometimes redundancy introduces new and exciting failure modes!
I have been well aware of the issue with two NTP servers disagreeing on time which is why across a WAN I'd never have less than four. The extra one is in case of a failures isn't available and the other three can vote on who has the right time. Assuming they aren't all depending on GPS that is (when I started doing this there were fewer than 10 public NTP servers in the UK I think it's better now).
It does seem lots of places are blocking ntptrace now. Not sure how you can abuse this information.
This should be a wake-up call for many product and solution vendors in the broadcast, cellular wireless and the emerging 'rush to market' /'Wild West' of IoT, as well as the obvious risks for navigation and emergency services.
The 'precision clock' that can be derived from GPS (and similar sources) is as precise as its system controllers want it to be - So not a reliable enough single source for any critical application.
I can't think of many things, and virtually nothing IoT, that requires any kind of high-level time precision.
Pretty much every computer network on the planet is more than happy being within five minutes of their other servers. Everything else is happy "to the second" to just get the clock time.
Anybody deploying tiny little IoT junk and expecting it to be anywhere near accurate just because they loaded up NTPd and pointed it at their GPS receiver is in for bigger worlds of hurt than a 13ms time difference knocking things out.
Fact is, if it mattered, you should have another source of that data and be able to compare, or cope if it suddenly jumps. The clock in your processor is probably sufficiently reliable to detect such jolts. That you don't bother to program for them and just slap in an NTP / GPS library is really the problem, not that GPS might jump a little.
Most IoT devices will just use a simple RTC and crystal occilator for timing. More than accurate enough for their purpose. If the designer really wants he can grab an accurate time of the network and pass it to the RTC a few times a day (AFTER checks and sanitation of the signal ofcourse. Don't want to be resetting your RTC to 12:00 every day at 12:42 or something)
Funny that: Digital radio, that can't report the new year's bells on-time due to the various processing delays, falls over when there is a timing glitch of tiny proportions, while analogue FM just keeps working as they never assumed precise synchronisation of transmitters in the first place.
Why does digital radio need synchronisation at all is my question?
Surely, if it's digital, the sensible thing would be to just make it a mass-broadcast medium like a multicast channel and put buffers into devices.
Then small jolts of interference barely matter as you're likely sending enough to recover the original digital signal, and a buffer capable of feeding out corrected data while it's still correcting the newest data. You don't need synchronisation. And the delay could be on the order of only a second.
I honestly do not understand why DAB is not just VoIP over 3/4G, with some kind of buffer and multicast-repetition of data, and a data format that allows large amounts of bit-error correction as well as detection.
There's a reason that the only DAB device I own came "for free" with my car. And I never use it. And it still has FM anyway.
> Why does digital radio need synchronisation at all is my question?
As I understand it, the transmitters need very very precise synchronisation as they transmit on a single frequency (well actually a set of frequencies), with the receiver picking up whatever signal is strongest. If they drift off then that creates sum and difference frequencies that cause a lot of interference and block reception. This can be heard on (for example) aviation frequencies near a busy airfield - if two stations transmit at the same time, everyone gets to hear a horrible wailing and screeching from the radio. Also, the actual bitstreams must be synchronised so that every transmitter is sending the same bit at the same time.
> I honestly do not understand why DAB is not just VoIP over 3/4G ...
For the same reason that broadcast TV isn't going anywhere quickly - broadcast like this is a highly efficient mechanism to get the same thing to lots of people (especially when many of them are mobile). Any Something over IP service just doesn't scale that well.
DAB is a single frequency network. If all transmitters start broadcasting each word at the same time then the signals constructively interfere even if there are multipath distortions, provided that the delay is not too long.
Higher bandwidth is achieved with multiple subcarriers rather than higher frequency modulation.
That way you can get better coverage, particularly for mobile users, for a given transmitter layout or the same coverage with a smaller transmitter network.
Yabbut... FM radio is carefully positioned such that adjacent transmitters use different frequencies, and require something like RDS to allow your receiver to switch automatically between transmitters.
DAB radio puts everything on the same frequency, and is designed to allow a 20% overlap of a 1ms packet, so you can get the same signal from more than one transmitter without interference (and with constructive reinforcement at the transmitter boundaries).
Different approaches to the same problem.
Seriously though, there's more than 24 satellites, with at least 3 or 4 in sight at any one time, and usually 9 or 10 visible. I know the GPS chips for cellphones are multichannel as you need to track multiple satellites for location.
So the $1500 clock chips don't do a sanity check across the visible constellation? Seriously? Especially since the timing is apparently so important?
How many functioning GPS satellites are there up there and how long would a non-functioning GPS satellite hang around over the Midlands ?
Methinks either something is very wrong with peoples timekeeping equipment or someone has not explained what exactly happened. There's a story there somewhere.
They're in 12-hour orbits. You wouldn't use one all the way down to the horizon, but you'd still get a few hours on any given satellite. The Fine Article says that "several other satellites were affected as a consequence".
You need to have a fix on at least four satellites; the times they arrive give you four equations, from which you get four variables (your position -- lat/lon/altitude or xyz) and the difference between your GPS device's time and the times on the satellites. I suppose the folks using GPS for timing _could_ (and maybe do) get their time from a single satellite: they already know where they are. (This sort of error should affect not only their timing, but their position. If their GPS told them they had moved when they knew full well that the device was in one place, that would have been a red flag.)
In general, there are more than just four satellites available at any given time (seems to me the GPS in my car -- an el cheapo device -- can usually get six or more). 13 microseconds corresponds to a four-kilometer error, and it seems to me that ought to stand out as "bad data". So the error should be quite noticeable, even without having your own atomic clock to check against. The software would, at the very least, notice discordant data. If poorly written, it would try to include the bad bits anyway and end up with an unusually large uncertainty value for your time and position. If a little better written, it might be able to figure out that one or two satellites appear to disagree with everybody else, and would toss their data.
As you say, GPS receivers work with multiple satellites and it seems, from the "several other satellites affected" report, that GPS satellites talk amongst themselves or to a ground station.
Anyone designing a GPS system (whether the satellite end or the receiver end) would consider the case where data from at least one satellite or ground station was in error. We are talking military on the satellite side, and high precision timing (for one use of GPS) on the receiver side so it is not a slap happy crowd of lazy engineers and programmers that produced these systems.
OK GPS is complex and a truly unforeseen or very low probability event might have just happened but, to me, it looks odd.
I'm curious, what telecoms companies rely on GPS for any timekeeping? Muxes and the like (at least the ones I saw) had no external clock/timing source, and POTS exchanges (even digital ones) have been around long enough to pre-date such clock sources so I can't see them relying on that either (I've seen a few 5ESS systems in the USA, didn't go into the details of the different inputs)
Mobile phone companies maybe?
"Lights out for Space Vehicle Number 23: UK smacked when US sat threw GPS out of whack"
This sentence don't make any sense ..
"DAB transmitters must broadcast at exactly the same frequencies and, in order to synchronise, they lock on to GPS satellite signals"
What happened when the BBC tested the DAB system for when the GPS signal goes out?