A Few Additional Technicalities On The Opera Measurement
    By Tommaso Dorigo | November 18th 2011 09:06 AM | 57 comments | Print | E-mail | Track Comments
    About Tommaso

    I am an experimental particle physicist working with the CMS experiment at CERN. In my spare time I play chess, abuse the piano, and aim my dobson...

    View Tommaso's Profile
    After a more careful reading of the paper, the listening to a seminar on the result, and some discussions, I can share with you a few more details on the Opera measurement.

    • The jitter: as you can see from the distribution I posted in the other article earlier today, showing the "delta t" for the 20 new neutrino interactions obtained from internal and external charged current interactions detected by Opera during the fast pulse runs of the CNGS beam, the jitter of 50 nanoseconds is causing the neutrino events to occur at random according to a uniform distribution, a "box" of width equal to 50 nanoseconds. What is causing this jitter had escaped me during the first careless and quick reading I gave to the paper last night. It is due to the fact that Opera cannot actually measure times with better than 50 nanosecond resolution: their "clock" runs at 20 megahertz, so they have a 50-ns "granularity" in the measurement.

    • This raises a rather annoying question: what would happen if this clock were read with just "one tick" of mismatch ? It might shift the delta t measurement to a more reasonable "8+-10 ns" - or to a unarguable 108+-10 ns. Which of the two is more probable, I let you decide.

    • The dataset: the new analysis of the 3-year data includes 5% fewer events than the original one. I had overlooked this fact earlier, but I had some strange feeling when I wrote "15223" in the earlier post. So what happens is that Opera discarded, upon reanalyzing the data, some 800 events which were found to be unsuitable for analysis: due to noise or otherwise unusable. Carefully analyzing your data multiple times helps!

    • An effect of the Earth rotation had not been taken into account in the original preprint: the trajectory of the neutrinos is longer by about 60 cm than what one measures on a fixed reference system. This, however, goes in the direction of making the measured speed even higher, albeit by just one thirtieth.

    • The light fiber which introduces a 40,000+-1 ns offset (which is subtracted along with other factors when computing the time difference) will be remeasured soon. The refraction index, and thus the time it takes for a signal to propagate in it, depends of course on the temperature, as well as on its stability as a material. At CERN a similar fiber was monitored continuously, showing no funny behaviour - so this might not be a problem in the measurement after all.

    • Now new measurements will be performed. One idea would be to unroll a light fiber from CERN to Gran Sasso, to bypass completely the GPS measurements. Another implies carrying physically an atomic clock from the point of creation to the point of detection of neutrinos... We will see what that brings.


    "This raises a rather annoying question: what would happen if this clock were read with just "one tick" of mismatch ? It might shift the delta t measurement to a more reasonable "8+-10 ns"
    This would need a very crazy conspiracy between the neutrinos and the clock if you also wanted to ensure the jitter, or not?
    No no, I am just talking of a mistake on the experimenters part. Ever tried to readjust your watch upon making an intercontinental flight ? Sometimes one errs by a full hour, but he was mistaken by just one bit!

    This is the most reasonable explanation I've heard so far. Unfortunately.

    I still believe that the jitter would look differently in that case, but lets not start looking too closely at a mere 20 data points. The second reason why I do not believe the 50ns binning is the systematic error:
    OPERA's previous results show the early arrival time to be 54 ns and 68 ns for 14 GeV and 41 GeV neutrinos, respectively.

    Well can you please find out whether I'm right? It would make much more sense in every possible way to timestamp the events than to try to run a stopwatch. You still can't rule out a really silly software bug but most off-by-one errors occur when you start and stop a counting loop: see marco's post. But it would be incredibly silly to have such a loop at all when there's (I damn well hope!) a perfectly good Master clock on site.

    "The light fiber which introduces a 40,000+-1 ns offset (which is subtracted along with other factors when computing the time difference) will be remeasured soon. The refraction index, and thus the time it takes for a signal to propagate in it, depends of course on the temperature, as well as on its stability as a material. At CERN a similar fiber was monitored continuously, showing no funny behaviour - so this might not be a problem in the measurement after all."

    The fact that a similar fiber did not show variations is not a gurantee by itself. There are individual variations in both refractive index and dispersion along and across the fiber that must be accounted for. The temperature needs to be carefully controlled, considering the huge potential offset in signal timing. Possible internal microcracks inside the fiber core can also be troublesome.


    Well, once again, it simply proves E=MC^2 is not correct. The solution is E sub t = M sub t L^2.

    Energy at a given point in space-time (with a variance of energy effecting it) is equal to mass at that point in space-time (same variant of energy effecting it) traveling at the speed of light squared.

    Any takes on that formula?

    Thanks for the update? Is the seminar recorded and available on the web?

    I'd be more than happy to walk the atomic clock from Geneva to Gran Sasso if anyone wants to pay my expenses!

    That's a pretty good publicity stunt.  How much money are you talking about?
    Watch out world, Hank smells a business opportunity... ;)

    Well, yeah, we can settle this business once and for all and make a fun Science 2.0 documentary out of it.
    It's a great idea, Hank!

    If really interested, you should contact INFN (I could act as liaison)- I think they are already thinking of doing this. Unrolling a cable is not such a big deal after all. Even less so carrying around an atomic clock...

    Brett says he will do it for expenses and I think it's an interesting idea.  
    I'll inquire.
    lol, How long is that cable?

    I'm thinking it's not going to be on a spool one man's going to carry, maybe an international effort is more appropriate :)
    Never is a long time.
    A couple hundred euros - or 15 minutes of fame may suffice!

    Tommaso, the point you raise about the 20 MHz clock might be the answer. It they have not already done this, the Opera team should have someone running through their software. It is very common bug to have a counter of by 1. A simple initialisation error can be enough. Also very common: forgetting to increase a counter at the last iteration of a loop. These problems can be very hard to spot. Certainly if you are not explicitly looking for them.

    "One idea would be to unroll a light fiber from CERN to Gran Sasso"

    Wait, are you telling me that, after all, Gelmini's tunnel will be really built? :D

    For sure it will be built.

    Politicians are just missunderstood visionaries.


    if an atomic clock were transported indirectly for such a distance how much time will you need to adjust for for the time distortion for the velocity.

    dont use a light fibre - try a laser in a vacume tube parallel also through the rock - the least controversial solution.

    A 730 km long vacuum tube dug 11 km below the surface ? I think you underestimate the issues.

    Since neutrinos have mass they cannot travel superluminally or even at the speed of light.

    You write "Another implies carrying physically an atomic clock from the point of creation to the point of detection of neutrinos..."

    The paper says

    "This time link was independently verified in 2011 by the
    Federal German Metrology Institute PTB (Physikalisch-Technische Bundesanstalt) by
    taking data at CERN and LNGS with a portable time-transfer device commonly employed for
    relative time link calibrations."

    I took that to mean that someone actually did carry an atomic clock from CERN to Gran Sasso, but maybe that's not what happened?

    This could work.... 
    Now new measurements will be performed. One idea would be to unroll a light fiber from CERN to Gran Sasso, to bypass completely the GPS measurements. Another implies carrying physically an atomic clock from the point of creation to the point of detection of neutrinos... We will see what that brings.
    ...and would be allot simpler than using GPS which seems to be the weakest link in their system.   So long as the optical path length and neutrino path length are the same we just have a straightforward race.  Whichever particle is detected first is the winner.

    I thought of something similar but more extreme way of testing this.  If we could place a satellite in geostationary orbit, or farther, with some source of both neutrinos and photons on it.  When the source is triggered it releases the photons and neutrinos as part of the same reaction.  Whichever signal we detect first must be the fastest.*  It would be a custom made supernova SN1987A.

    *An atom bomb would do the trick...but no one would ever let us do that.  
    Science advances as much by mistakes as by plans.
    Hi Hontas,

    it is not a race between photons and neutrinos. The light fiber would allow to synchronize the clocks at CERN and LNGS, once the transit time on the fiber is known.

    Ah I see.  That's still an improvement over relying  on GPS.  I hope they really do it. 
    Science advances as much by mistakes as by plans.
    >> what would happen if this clock were read with just "one tick" of mismatch ?

    there could be a bias if an event is recorded with the latest tick of the clock (and not the next one yet to come).
    on average this would shift the time stamps by 25ns and account for half the effect.
    do we know how the clock works at the CERN side (when the pulse is generated) ?

    on the other hand, this is so trivial that i am sure they have thought about it quite a bit (pun intended!) ...

    Hi Wolfgang,

    at CERN the clock works at higher frequency -no jitter from there. But about the GS clock: there was no mention at all of the 20 MHz clock in the previous Opera paper - I let you draw your own conclusions about it, cannot comment further.
    So trivial as it might be, I would not bet my balls on it being correct.

    One possible replacement for GPS in OPERA would be to use the RF carrier frequency from a geosynchronous satellite as a source of clock pulses (after differentiation to create discrete pulses). One candidate might be downlink signals from a direct TV broadcast satellite; in Europe the Astra satellites operate in the Ku band (10.7 to 12.75 GHz ). Assuming compensation for distance and fiber corrections, atmospheric delays, and receiver and other circuitry, this could potentially provide a continuous ~8 ns. clock pulse to both sites, less than the claimed accuracy of <10 nanoseconds ns. ( Aligning the clock pulses at CERN and Sesso could be done (in non-real-time) by matching  modulated video signals in recordings at both locations. However, I do not have data on satellite signal carrier frequency jitter, which needs to added to the 8 ns. 

    Note that this solution, while eliminating putative errors due to GPS movement, would not be as precise as the 2.3 ± 0.9 ns. available with the OPERA system (to correct the nominal 100 ns precision of GPS, PolaRx2e receivers achieve a claimed difference between the time base of the CERN and OPERA of 2.3 ± 0.9 ns. (according to T. Feldmann, “Relative calibration of the GPS time link between CERN and LNGS, Report calibration CERNLNGS 2011”, OPERA public note 134, 2011, (Of course, subsystem measurement precision does not imply system measurement accuracy.) 

    I would appreciate hearing any critiques of this speculative concept.  

    Hi Amara,
    thanks for your interesting note. I am totally unfit to have a meaningful discussion on this highly technical issue. I however feel that if this system can be used together with the normal one, the two would cross check one another quite easily. Have no idea how hard it is however. One piece of systematic uncertainty that would remain is the transfer of the signal down to the Gran Sasso mine, done with a light fiber which is 8km long.
    I hope other readers here can comment on the matter I'm more than happy to stand on a side and take notes.
    If people are interested in the CERN-LNGS time transfer and are thinking about alternative methods to double-check the GPS common-view scheme: FYI, there is a wiki with all the information about this time transfer:
    And there is also a mailing list for those who would like to ask questions or propose new schemes:
    A fiber link and a traveling atomic clock seem indeed to be in their agenda.

    Xano Jsp
    I have a question regarding the methodology taken by the Opera collaboration that I hope some of you may be able to answer.

    Could you explain how OPERA's methodology compensates for the relativistic effects of site-to-site time measurement changes, due to Earth-Moon-Sun interactions and differential gravitational tidal forces in particular?

    More succinctly, have they accounted for differential relativistic tidal forces? This is not to be confused with newtonian tidal forces that do not compensate for time dilation effects.

    Many thanks,


    Hi Martin,

    short answer: no.
    Hi, I'm just an interested layperson, reading about this in my lunchbreak, so its probably been discussed a thousand times already, but could a 18m baseline discrepancy be caused an incorrect model of the curvature of the earth, eg a fault in the geospatial projection, or an assumption that the earth is round. This would mean that althogh the locations of the source & receiver on the earth's surface are correctly pinpointed in terms of latitude and longitude, and presumably elevation relative to an average sea-level, the distance may be less than caluclated. Or does the GPS provide absolute x,y,z locations independent of the earth's curvature?

    Hi Stuart,
    the position has been determined with very good accuracy and there is no reason to believe in such a huge error. The timing measurement is probably affected by other issues -the determination of the transit time of light in the 8km long fiber,for instance.
    Wouldn't measuring traversal time of light on an optical fiber reintroduce complex statistics. Light travels in rather random fashion through fiber, and probabilistic analyses are required to get the distance it travels.

    Last reflection that I had on the neutrino anomaly

    could the anomaly be due to an additional laggard decay on the hadron stopper?

    18 meters (3 graphite 15 iron) =60 ns

    It wouldn't be as nice as a vacuum but those that survived, survived.

    Specially with the additional extraction made by the first 3 meters of graphite of the Hadron stopper

    It might be visible on the neutrino detection form:

    - two normal functions instead of one,
    - even, one might dream, neutrino interference patterns

    And the effect could be different on different sized beams.

    - larger sized beams could give bigger signals

    It might explain the difference between higher and lower energy neutrinos anomaly

    - kaons still decaying in the stopper
    - freshly made kaons, graphite stopper born, decaying fast in the stopper and beyond

    As the TOF is compared on a form fitted PDF, with an additional 60ns of true extraction, assuming that the last to arrive were the last protons, we'd assume the first had arrived 60 ns earlier.

    At this point I refuse to discuss any explanation of the 60ns anomaly until new data on the light fiber time delay measurement arrives. I am convinced that the uncertainty on that 40000 ns delay cannot be 1ns (MINOS has a much larger uncertainty on a similar fiber, and they measured it with four methods, finding four different values, while Opera did it only with one method!).

    As you wish, I just thought you might be interested and check it or just mention it to someone who might (as you're in CERN)

    An 18 meter (60 ns @ C) hadron (cost-)stopper that acts as a second target beyond focal point.

    Thank you anyway

    Ummm, the four different MINOS values could well be from the low quality of the measurement techniques rather than innate variability in the fiber delay. And even if the fiber delay truly had more variance, it is hard to see how both MINOS and OPERA ended up with mean superluminal values - more variance implies at least one should have seen subluminality.

    The EXFO OTB-300 OTDR, which uses reflections off a terminated but not looped-back fiber cable, claims an accuracy of 1m +/- 0.0025% of distance. The OWL BOLT-NL, which runs a true loopback test, claims an accuracy of 0.01%. The Agilent E6000-91031, another OTDR, also has an accuracy of 0.01%. Agilent, in particular, has high credibility in the electronics world. OPERA's reported 0.0025% with a custom setup is not out of line with what can be achieved with off-the-shelf equipment. The question seems to be more why MINOS alone has such trouble with fiber cables.


    Please refer to this page for any questions on the timing calibration at CERN.

    The fiber calibration at LNGS was made with the same methods already used at CERN, a travelling cesium and a two ways calibration. The travelling cesium was affected by the instabilities of the LNGS GPS at that time. Both calibrations agreed at the ~13ns level.

    Kind Regards,


    Interesting Ajoy. However what is known about the time and temperature stability of the transmission time of these fibers ? Is that, too, known with such splendid accuracy ?

    I don't think so (typically, length measurements are done to detect fiber faults, shorts and breaks). But aren't they remeasuring the fiber delay? That should address the stability issue easily.


    if you unroll a fiber from CERN to Gran Sasso, the light in that fiber will travel slowly that c.
    It will be not in emply space but in glass. How can this be even teoretically useful?


    Hi Anon,

    the fibers exist already -the issue is getting exclusive use of one of them for the required timing measurements, from the telephone company who owns it.

    There is no need for light to travel at the speed of light in the fiber. In fact, it could be ants traveling from CERN to LNGS. As long as we know that they take always the same time to travel from there to here, we have a perfect tool to synchronize the measurement. We just need to ask the ants to travel backwards at the same speed once, to measure the travel time with one very precise clock (by dividing in half the two-way ride time.

    But the ants need to go back and forth at the same speed, from here to there, and from there to here. It isn't clear that holds for fiber over 700 km, with repeaters/amplifiers/etc., unlike line-of-sight laser. If the delays are asymmetric, there really is no way to separate the clock asynchronicity from the fiber delays. One could perhaps average over many fibers, assuming the asymmetry is '+' and '-' with equal chance. Why don't they move the clocks to the same place (slowly or whatever) and recheck sync?

    Yes, carrying an atomic clock seems a good option.
    I don't know why, but I'm literally obsessed with this fact of the delta-speed of Opera's neutrino being very close to the orbital velocity of GPS satellites.
    My question is: does anyone know if the guys in Opera have looked for some correlation between neutrinos detection times and arrangements of GPS satellites hovering over LNGS ?
    Finding some "regularity" of that kind would hint to some relativistic effect not properly taken into account, wouldn't it ?

    Yes and no... Individual detection times are possible only in the latest set of data (20 neutrino interactions), and even then, there's a 50ns smearing due to the 20MHz clock. I doubt you'd find anything useful there..

    I think my comment was not really clear (no wonder, as I’m pretty confused myself)… I try to explain what I mean.

    The “rationale” is that if they failed taking into account all facets of SR, that might show up as a correlation of the flight time of neutrinos with some parameter of the GPS satellites (mean distance, velocity vectors…whatever). Or maybe not, but it should quite easy to give it a try,,,

    And yes of course they can do that only with the 20 latest events, but:

    1) If a systematic error is actually hiding there, the correlation could emerge even with such small statistics (and that is going to grow anyway)
    2) Hubble’s law was first discovered on a very small (and partly inconsistent) data set. :-)


    Hello Tommaso,

    I guess that the possibility that the two clocks are affected by relativistic local aspects (different places -> different gravity, different local mass concentrations) is under consideration.

    What about to move the clocks close one each other and trigger a free running start, maybe at CERN, then moving one back to the Gran Sasso labs letting him to free run & counting for a certain amount of time but long enough to show up possible time shifts, then move it back to CERN to trigger a common “count stop”.

    I wonder if the two clocks finally will count the same amount of ticks….

    Thank you
    Silvano Mezzetti

    Ciao Silvano,

    this and similar ideas have been proposed by many. We will see what will be implemented next. The recent run with short pulses has on one hand canceled some potential sources of systematics, but has revealed some minor defects that the new analysis has improved. The situation is still evolving.

    What if the speed of light isn't constant, as some physicists suggested recently?

    What if the speed of light varies through time and space?

    That would create some interesting theory. At least I think so.

    Antimatter is the mind and consciousness of all living entities.

    You are your own universe.

    Reality is where the minds (antimatter) meets the physical universe.

    Interested? Then read my philosophical multiverse theory.

    Google crestroyer theory, and find it instantly.