Neutrinos: Still not quite superluminal
February 22, 2012 1:07 PM   Subscribe

Did you try unplugging it and plugging it back in? Last year's faster-than-light neutrino observation may be explained by a loose connection between a GPS unit and a computer. posted by 0xFCAF (64 comments total) 6 users marked this as a favorite
 
as predicted
posted by exogenous at 1:09 PM on February 22, 2012 [1 favorite]


Not to come off as too big an ass, that's the sort of mistake I would have made myself. If it's plugged in, it's plugged in, right?
posted by exogenous at 1:10 PM on February 22, 2012


Or maybe that's what the time travelers from the 26th Century who rigged this here fake loose connection want us to think? You know, to fool us into abandoning this research, thereby avoiding some unspeakable future disaster that we'd have brought upon ourselves?
posted by .kobayashi. at 1:11 PM on February 22, 2012 [10 favorites]


Aw, bummer.
posted by Xoebe at 1:11 PM on February 22, 2012


*facepalm*
posted by kmz at 1:12 PM on February 22, 2012


They're really regretting not buying the Monster brand GPS <-> Neutrino Computer cable.
posted by 0xFCAF at 1:12 PM on February 22, 2012 [57 favorites]


Einstein is laughing.
posted by lee at 1:12 PM on February 22, 2012 [1 favorite]


kobayashi, nobody bought the Meddling Time Travelers excuse when you were three hours late for Susan's party, so could you please drop it?
posted by theodolite at 1:13 PM on February 22, 2012 [12 favorites]


The neutrinos went back in time to break the connection, so that we would never know their true nature.
posted by Sticherbeast at 1:13 PM on February 22, 2012 [1 favorite]


The neutrinos went back in time to break the connection, so that we would never know their true nature.

I already said that a few minutes ago.
posted by swift at 1:14 PM on February 22, 2012 [1 favorite]


I already said that a few minutes ago.

I'll preemptively address this concern later.
posted by Sticherbeast at 1:15 PM on February 22, 2012 [11 favorites]


This perfectly illustrates poor troubleshooting by even smart people. I work in IT, and this sort of thing makes me want to kill myself roughly on a daily basis.

This should be the focus of a mandatory class in all schools, at all levels.

My first assumption about this, which I commented on the other post, was "The distance is likely not measured properly." This could have been verified and corrected before anybody made fools of themselves releasing news of FTL particles.
posted by Threeway Handshake at 1:16 PM on February 22, 2012 [2 favorites]


I think it's too early to say that this is the final and definitive explanation of these results. We've seen a few "and so it was all a simple experimental error" reports come out before now and then be shot down later. This might very well turn out to be genuine, but so far it's just a new hypothesis.
posted by yoink at 1:17 PM on February 22, 2012


From what the short article says, I think they were measuring the distance fine, but they screwed up the measurement of the time.

It's too bad there was so much matter in the way. Otherwise they could have just verified their experimental setup by timing the same trip for light.
posted by aubilenon at 1:19 PM on February 22, 2012


My first assumption about this, which I commented on the other post, was "The distance is likely not measured properly." This could have been verified and corrected before anybody made fools of themselves releasing news of FTL particles.

And your first assumption, if this story turns out to be correct, will turn out to have been wrong.
posted by yoink at 1:20 PM on February 22, 2012


Pretty much anybody who knew anything about physics said the same thing about ftl neutrinos: "Not impossible, but incredibly unlikely to be correct."
posted by empath at 1:20 PM on February 22, 2012 [1 favorite]


theolodite -- Everyone was wrong at Susan's party too. I mean the old "If there were time travelers, wouldn't they have prevented the hors d'Ĺ“uvres from burning?" bit, really? Well, maybe they like their pigs-in-a-blanket blackened in the future.
posted by .kobayashi. at 1:20 PM on February 22, 2012


That's so glossed over as to be unusable. A loose fibre connection causing a 60ns delay? Light travels in vacuum (and, for all intents, in air) 30cm/ns, or 1800cm, or 18 meter, or nearly 60 feet in old money.* A loose connection is not adding 60ns to the signal path. It'll certainly drop the level by a few dB, but it's not adding a delay. The distance where refractive effects are a concern are tiny -- we're talking picosecond delays. You are not getting 60ns of slop with a loose connection.

You could, however, get 60ns of slop if you incorrectly measured the delay in the first place, and I could easily see 15m of fibre between the time source and the instrument, given that the speed of light in your typical fibre is about 70% of vacuum, so, you'd travel 30cmx.7=21cm/ns, or 12.6 meters, depending on the exact material and if it's multimode or singlemode.

So, IMHO: This loose fibre connection is nonsense. If it turns out that they mismeasured, or forgot to factor in the speed of light through the fibre link, that I will certainly believe, and 60ns delay for 15m of fibre is certainly close enough to match.

And, guess what? 15m is a standard length for pre-terminated fibre cable.




* Oh, that's right, a foot a nanosecond.
posted by eriko at 1:22 PM on February 22, 2012 [10 favorites]


This is good news for all of us: next time I make an embarrassing mistake, I can console myself with the thought that at least my error didn't receive this much worldwide publicity.
posted by fredludd at 1:26 PM on February 22, 2012 [6 favorites]


The article is basically unsourced rumor, too..
posted by empath at 1:26 PM on February 22, 2012 [1 favorite]


The physicists on social media have been reposting this all day, mostly saying "does anyone know anything about this?" No one has confirmation yet, so we're all waiting for that before declaring this case closed. Right now, this is just rumor and speculation. If I hear more, I'll fill you guys in.
posted by physicsmatt at 1:26 PM on February 22, 2012 [3 favorites]


Occams Razor is great--but it doesn't mean "if it doesn't conform to my preexisting expectations then it is false, and if it does, it is true."
posted by yoink at 1:30 PM on February 22, 2012


They're really regretting not buying the Monster brand GPS <> Neutrino Computer cable.

I sense an upcoming collaboration between Drs. Dre and Hawking.
posted by griphus at 1:31 PM on February 22, 2012 [3 favorites]


In other Physics corrections news, it's now more accurately called the Higgins Bosom
posted by evilmidnightbomberwhatbombsatmidnight at 1:38 PM on February 22, 2012


So, IMHO: This loose fibre connection is nonsense.

Presuming that something wrong with the connection is the cause, could there be some sort of phase problem with the signal, i.e. that the loose connection caused the receiving end to miss some of the pulses entirely and when it "reacquires" it, it catches on to a segment of the signal that is 60ns off? I'm assuming several things about how it detects the signal, the type of signal that goes down the fiber, and how the system knows it's locked onto the correct time, but it doesn't sound like nonsense to me. I've seen stranger things in labs.
posted by chimaera at 1:39 PM on February 22, 2012 [1 favorite]


Mmmm...pigs-in-a-blanket...

*drool*
posted by darkstar at 1:41 PM on February 22, 2012


And I still consider a loose connection to be far more likely than superluminal neutrinos.
posted by chimaera at 1:41 PM on February 22, 2012


but it doesn't mean "if it doesn't conform to my preexisting expectations then it is false, and if it does, it is true."

Nobody ever said otherwise.

And your first assumption, if this story turns out to be correct, will turn out to have been wrong.

This has been written up everywhere in a number of ways, and it is still pretty unclear as to exactly why this "cablegate"error factored in to this. Regardless, the troubleshooting/verification of this GPS measurement -- which would be likely be the first thing to go over, would end up with the proper course of action, which is to correct for this error.
posted by Threeway Handshake at 1:43 PM on February 22, 2012


Occams Razor is great--but it doesn't mean "if it doesn't conform to my preexisting expectations then it is false, and if it does, it is true."

Occam's Razor has nothing to do with why people are doubting the result. One very good reason to doubt it is that we don't detect neutrinos from supernovas before we see them.
posted by empath at 1:50 PM on February 22, 2012 [4 favorites]


Science, a harsh mistress indeed.
posted by Chekhovian at 1:54 PM on February 22, 2012


There should be a rule or law for this. When it comes to computers and networking, any mysterious problem can be traced back to a faulty cabling problem. That's been my continual experience for 30+ years. It's the most simple answer, and the last one anyone thinks to check, and the one most likely to be true.
posted by stbalbach at 1:54 PM on February 22, 2012 [4 favorites]


Presuming that something wrong with the connection is the cause, could there be some sort of phase problem with the signal, i.e. that the loose connection caused the receiving end to miss some of the pulses entirely and when it "reacquires" it, it catches on to a segment of the signal that is 60ns off?

The usual method for this sort of high-precesion clock sync is called "pulse per second", though in modern time, it's usually pulse per microsecond. Basically, the clock source sends a plus on a known interval, and the receiver uses a PLL with that pulse to sync time.

The usual pulse lengths are far enough apart that a 60ns error from a missing pulse is very unlikely. It's easy to get microsecond sync from a pulse-per-second source, so a pulse-per-microsecond would be more than enough to get a +-1ns resolution at the destination clock.

What I suspect happened was this: They saw actual dropouts on the clock signal, and went to check the cable. Since this cable is timing critical, they would have checked to see the recorded installed length, so if they had to replace it (they didn't, it was loose), they'd get the right length of fibre and not wreck the calibration. Often, you have to go a long way to get to there, so you grab the spare before you go. But, hmm, the length of cable wasn't recorded. That's odd, but hey, mistakes are made.

Now, on a nanosecond scale clock, the pluse-per-microsend source has a very definite delay. You take this into account by putting an offset into the receiver. You measure the cable's delay time, and tell the receiver that the PPS signal will be that many micro- or nanoseconds late, and it can then sync to the far clock correctly. I'll bet they looked there, to see how roughly long the cable was, and....the offset is zero?

(faceplaming ensues)

Then, of course, to be sure, they grabbed a TDR, went out, wiggled the cable and saw the signal come back to normal, and then measured the delay product -- 60 ns.

(facepalming, again)

And, describing this to a reporter, the reporter took away "loose cable, 60ns" and wrote "a loose cable was causing a 60ns delay." And when they read that description?

(facepalming FOREVER)

Insert appropriate JPEG here.
posted by eriko at 2:13 PM on February 22, 2012 [15 favorites]


Well, it proves the system works. Science wins again.
posted by Blazecock Pileon at 2:17 PM on February 22, 2012 [5 favorites]


It's my memory that the OPERA results were presented with a big disclaimer that they were almost certainly wrong and probably would be independently falsified.
posted by CBrachyrhynchos at 2:32 PM on February 22, 2012 [2 favorites]


Thanks, eriko, that makes much more sense. *sigh* science journalism.
posted by chimaera at 2:32 PM on February 22, 2012


Or maybe that's what the time travelers from the 26th Century who rigged willan on-rig this here fake loose connection

Fixed that for you.
posted by Mr. Bad Example at 2:35 PM on February 22, 2012


A detailed announcement is scheduled for Thursday, followup tests scheduled for May. Sounds like someone broke an embargo.
posted by CBrachyrhynchos at 2:37 PM on February 22, 2012


Well, there's a shocker.

As I said at the time, I would basically have bet all my worldly goods that neutrinos were not travelling faster than light.
posted by Decani at 2:40 PM on February 22, 2012


(As a complete aside, eriko, "a foot per nanosecond" is the standard expression I was taught for semiconductor devices and circuit design. It's a very convenient human scale, and it always reminds me how freaking small a nanosecond really is! Even light - light! - travels only one foot in that time.)

I do sympathize with the OPERA guys, but it's like everyone said right at the start - it was always a rather far-fetched result. And to be fair, they were the first to say that they didn't really believe v>c either.
posted by RedOrGreen at 2:49 PM on February 22, 2012 [1 favorite]


*watches FTL Neutrinos fade from an old photograph*

Oh no... Doc, we're too late...
posted by Uther Bentrazor at 3:15 PM on February 22, 2012 [4 favorites]


It's funny, RedOrGreen, I always thought about a foot per nanosecond more as a limit on how slow light goes. Nanoseconds are tiny, sure, but people talk about femtosecond laser pulses and whatnot-- in which time light hasn't really gone very far at all.

Also the fact that I'm c. 6 nano-lightseconds tall means that no matter how hard I try, I cannot possibly respond to a request to "do something this nanosecond". Sorry, my feet haven't heard about it yet, they're too far away.
posted by nat at 3:28 PM on February 22, 2012 [5 favorites]


Sort of confirmation. New tests will be run in May
posted by physicsmatt at 3:46 PM on February 22, 2012


I sense an upcoming collaboration between Drs. Dre and Hawking.

Please, that's old news...
posted by Chekhovian at 3:47 PM on February 22, 2012 [1 favorite]


At least I could somewhat follow this breakthrough with my limited understanding of physics, too bad it was mistake :/
posted by opi at 4:13 PM on February 22, 2012


Measuring time is like writing crypto. It's really easy to get an answer but bloody difficult to be sure you've got the right answer.
posted by Skorgu at 4:15 PM on February 22, 2012 [1 favorite]


OPERA official statement reported here.
posted by physicsmatt at 5:37 PM on February 22, 2012


Occam's Razor has nothing to do with why people are doubting the result.

Quite. My argument is with those who claim it does.

Similarly, Occam's Razor has nothing to do with the reason that almost everyone in this thread is hailing this fragmentary report of a possible, unconfirmed equipment error in the original experiment as utterly conclusive proof that their suspicions were correct all along.

It would have been foolish to take the original experiment as "proof" of superluminal neutrinos. It is equally foolish to take this almost information-free report as "disproof." All we know today is what we knew yesterday, that the CERN team had some anomalous experimental results that required further examination.
posted by yoink at 6:08 PM on February 22, 2012 [1 favorite]


"Okay, okay, who's the wiseass who put the USB adapter on the end of the GPS Ethernet??? Albini? Anybody seen Albini??"
posted by Twang at 6:55 PM on February 22, 2012


Now, on a nanosecond scale clock, the pluse-per-microsend source has a very definite delay. You take this into account by putting an offset into the receiver. You measure the cable's delay time, and tell the receiver that the PPS signal will be that many micro- or nanoseconds late, and it can then sync to the far clock correctly. I'll bet they looked there, to see how roughly long the cable was, and....the offset is zero?

Sounds about right. FTA:

After tightening the connection and then measuring the time it takes data to travel the length of the fiber, researchers found that the data arrive 60 nanoseconds earlier than assumed.

So the fault wasn't a faulty cable, but it did cause them to test the length of time for packets to traverse the fibre and their previous offset was 60ns out. Whoops. Still, they themselves said it was almost certainly experimental error, and they were publishing their results to try and figure out where it was. So yay science!
posted by ArkhanJG at 7:01 PM on February 22, 2012


Re "incorrect cable length" hypothesis...

1) I thought the exciting part of the OPERA results was that some, not all, of the neutrinos appeared FTL. Wouldn't changing the offset just change the position of the computed population distribution in time, but not the shape and limits of the distribution itself, i.e. wouldn't the effect just be that fewer neutrinos, observed or not, are FTL?

2) It would seem reasonable that the whole apparatus should have a calibration mode and signal path for empirically measuring and compensating for offsets and higher-order effects prior to the experiment being run. Does OPERA NOT have this, and if it doesn't, is this a responsible way to conduct such an experiment?
posted by ZenMasterThis at 7:16 PM on February 22, 2012


Also, what eriko said. 60 nsec is a metric assload of time to be explained-away by a loose fibre optic connection.
posted by ZenMasterThis at 7:25 PM on February 22, 2012


ZenMasterThis, the experiment originally couldn't measure neutrino-by-neutrino timing. So what they saw was an overall off-set that could be interpreted as FTL neutrinos. If you look at my previous posts on this, it was possible in the first configuration that none of the neutrinos were moving FTL and their timing was correct, BUT there was some issue with how the beam of neutrinos were being produced which screwed with their timing measurements. This is perhaps where you got the impression that not all of the neutrinos were moving FTL.

In the second set of experiments from OPERA, they shortened the neutrino pulse width, so they were closer to measuring individual neutrino speeds. This removed the possibility that what we were seeing was some pulse-shape effect, and made the best interpretations EITHER that (nearly) all of the neutrinos were FTL or that the timing circuit was wrong. As far as I was aware, this effectively removed the only "particle physics" explanation of the anomaly that wasn't a wild violation of relativity. At this point, the most straightforward explanation if their timing was right was that ever neutrino was 60 ns fast.

As for 2, I'm not sure what you mean. It's a little hard to have a calibration mode when you are already using the most accurate clocks in the world to make your measurement. If you want to calibrate them, what do you use? More accurate clocks? If you have those, you use them for the experiment. Also, keep in mind that you can't move the clocks together to compare, partly because some of the clocks are in space, and partly because moving the clocks introduces non-trivial GR corrections that you want to avoid.

But more generally, yes, these experimentalists did do (nearly) everything possible to check their experimental apparatus before running and then again after getting their crazy result. As embarrassing as this explanation is (if it turns out to be correct), it shouldn't be taken as evidence of lazy scientists, but as a sign of how hard complex modern science is to do, how amazing it is we can get the results we do get, and why scientists are so hard-nosed about experimental results that don't match the theory. We all expected this result to be wrong, and I think we will find out we were right in the end. But, to be clear, this isn't the end yet, as we have no confirmed statement from OPERA on what went wrong, and even if this report is true, we will still want to see MINOS upgrade their timing to get a 2nd, independent opinion.

However, if you look on arxiv, you will see a LOT fewer theory papers on this, even before the results are made public (presumably in May, from what I hear), because, though we all acknowledge that there is still a possible anomaly, we all get to choose what to work on independently, there's lots of pressure to do "interesting" things, and so theorists will follow their nose and work on something that is more likely to be "real" new physics. That's just how science gets done.

Overall, everyone paying attention just got an object lesson in how particle physics (and really, most any science) works. There was a crazy result that would overturn a century of established thought. We scientists all doubted it, including the people who did it. Everyone pointed out what could be wrong. Some theories were written down to try to explain what would have to be true if this crazy result wasn't a fluke. Major renovations to the experiment were done, and failed to "fix" the problem. Further explanations were proffered. Dozens of people spent months crawling around on their knees behind desks rechecking connections. Some possible errors were found, and if they pan out, we'll all go "told you so" and move on with a few more jokes about the wacky physics in Gran Sasso. The only thing different here was that it was Einstein's theory, and so there was a much more public spotlight on the whole process.
posted by physicsmatt at 7:42 PM on February 22, 2012 [3 favorites]


Thanks, physicsmatt. I have a little background in mixed-signal test instrumentation and network timing. What I meant was in complex instrumentation systems there's usually an end-to-end self-test mechanism that permits sanity-checking the instrument against itself before introducing the signal under test. Yes, there is always the problem of "how do know how good we are when we're already using state-of-the-art references." The solutions usually involves trade-offs between speed, accuracy, and other parameters, and they often involve building calibration tables with coefficients for a compensating function which fits a theoretical model of the apparatus.

In this particular case the whole apparatus is spread-out over hundreds of km (not including reference clocks), so I'm clearly out of my depth. I wonder if they considered leasing a dark fiber pair between the two locations, if that was even a possibility. I can imagine using a TDR to determine the exact length of the dark fibers, then using a master / slave retiming loop between the two sites to calibrate or even conduct the experiment itself. Could've potentially eliminated the need for the GPS link altogether.

Well, they're really, really smart guys, much smarter than me, so I'm sure they'll get it all sorted-out eventually!
posted by ZenMasterThis at 8:13 PM on February 22, 2012


A graphic on Wikipedia reports that the distance between GPS receiver and the master clock at San Grasso is 8300 m. An independent description of the San Grasso Neutrino lab reports that it has an average of 1300 m rock cover. (It's apparently built into the side of a highway tunnel.) The latency calculations on Wikipedia show that they're compensating for approximately 16000 ns of equipment latency (which doesn't include cable length). So yeah, it's not as simple as a 15 m patch cable.
posted by CBrachyrhynchos at 8:28 PM on February 22, 2012


Ars Technica is reporting that there are two issues with the timing that have been uncovered, one making it faster, and one slower. One does indeed involve an optical cable.
posted by Canageek at 9:24 PM on February 22, 2012


I do kind of feel sorry that the story ends this way for the team who could've gone down in history for their discovery. Well, at least they had their 60ns of fame.
posted by TreeRooster at 9:37 PM on February 22, 2012


I think we should be patient and let the OPERA team make a conclusive statement before everyone breaks out in their little I-told-you-so dances. The actual statement on the OPERA web page is more like "we found a couple of errors and we're reviewing our results - more to come." I for one don't pretend to have an instinctive understanding of neutrinos and will need to take the experimentalists on their word in this matter.
posted by newdaddy at 10:08 PM on February 22, 2012 [2 favorites]


I saw Steve Myers, Director for Technology at the LHC, speak a few months ago and he was asked about this. He said when the group came to him with the discovery to get his sign-off to release it to the press, he suggested it was probably a cable error. They said it wasn't, they had checked. So he said, 'Well go ahead, but you guys are going to look pretty silly if it turns out to be a cable error.'

I can see why he's Director for Technology at the LHC, and I am not.
posted by StephenF at 2:29 AM on February 23, 2012 [3 favorites]


So, IMHO: This loose fibre connection is nonsense

Eriko, of course we don't have all the details yet, but the loose fiber doesn't strike me as completely crazy. Your analysis, which is quite correct, assumes that this fiber is carrying a single pulse indicating the triggering of some event. But this might not be the nature of the signal sent through this fiber at all. Instead it could be a digital "heart beat" or synching signal made up of amplitude modulation, or phase shift keying, between two devices. It could be a frequency reference between the atomic clock and some counter. A "loose fiber" might actually mean a slightly damaged cable, leading to coupling into a cladding mode (which can involve much more significant delay), or worse, an etalon, leading to multiple pulses at the receiver!

Maybe the best analogy would be an improperly terminated 50-Ohm rf cable, from which I've seen otherwise sane instruments driven completely crazy: double-counting of pulses, faulty logic, timing jitter, reflections down the cable that corrupt a healthy signal, smoking amplifiers, etc.

In other words, it's maybe crazy enough to be a nice explanation. And it sounds like just the sort of thing that has ruined my week several times in the past, leaving me ever more paranoid, twitchy, and an utter bore to uninterested parties.
posted by fatllama at 6:54 AM on February 23, 2012 [1 favorite]


It would be hilarious if of the two possible errors OPERA have discovered, the one that turned out to be real was the one that made the neutrinos arrival even faster.
posted by yoink at 10:53 AM on February 23, 2012


The system depends on accurately subtracting over 8,000 ns of signal-processing time from every timestamp. (My earlier post was wrong.) Anything that increases the signal-processing time is going to affect the calculated timestamp.
posted by CBrachyrhynchos at 11:50 AM on February 23, 2012


In human terms, it's like trying to timestamp your email to the minute using postcards from Greenwich that are delayed by 5 days.
posted by CBrachyrhynchos at 1:37 PM on February 23, 2012


Bit of a round up: http://profmattstrassler.com/2012/02/24/everybodys-a-critic/
posted by physicsmatt at 4:19 PM on February 24, 2012 [2 favorites]


I only just found this thread because I only just found out about MyMefi, which turned it up after I had tuned my preferences. Anyway...

One of the interesting things to me here is that if we had a problem like this, but with a result which *confirmed* expectations, we would likely never have found it.
posted by philipy at 10:40 AM on February 27, 2012


« Older Exclamation marks up the minister's arse.   |   Dementia be Damned Newer »


This thread has been archived and is closed to new comments