Why Syncing Atomic Clocks is Virtually Impossible | Judah Levine on UTC transcript

Let's suppose I have a clock and I want to tell you what time it is. Then I'm going to send a signal to you which says it's now 2:00. But when you receive it, you don't receive it at 2:00. You receive it by some time later which has to do with a transit time. A clock on the moon as observed from the earth would be running at a different frequency. That is if you observed the clock on the moon and you observed it from the earth, you would see the frequency is different. >> Is it the same on the Mars? >> No, Mars is different because Mars has a different gravitational potential.

Also, there will be Doppler shifts because the distance between the Earth and Mars is changing. So, the signals are very complicated because there are general alistic effects because the gravitational potential is different. But there are also Doppler shifts because you're observing on a moving platform from a moving platform. This week, we spoke with Judah Lavine, the legendary NIST physicist, often called the father of internet time.

For over five decades, he's helped define the seconds that run GPS, financial markets, and the internet itself. In this episode, we dive into what time really is, how we measure it, and how to approach the impossibly hard problem of synchronizing atomic clocks across the world. Please welcome Judah Lavine. UTC uh to answer your question, uh UTC is a paper time scale that is it is it is a scale that is it does not exist as a physical clock. uh is computed by the international bureau of weights and measures by the BIPM uh in French uh based on the data from a large number of clocks in a large number of laboratories.

There are about I don't know four or 500 clocks uh in uh in the major laboratories uh pretty much every country has a timing laboratory and the timing laboratory maintains a number of clocks and the BIPM computes the ensemble average and that's UTC. Now uh the the uh originally UTC I mean actually there are a number of of um of intermediate products before you before you get to UTC UTC is kind of the end product um the other products are the one that's most most widely known is called international atomic time TAI um and um uh it is also computed by the BIPM based on the same the same data u uh the current version of UTC uh dates from 1972 two uh when uh the current version I mean UTC existed before that but it existed in different definitions um and the the UTC time scale now is derived from international atomic time that is it's it's it's it's computed from international time but it has the addition of what's called leap seconds and the leap seconds are added periodically so that uh UTC does not deviate by more than 1 second from uh uh Greenwich meantime which is which is the older astronomical time scale. uh the the uh since 1972 uh there have been there were 10 seconds added to UTC uh to to maintain this link between uh UTC and and uh and the atomic the astronomical time scale which is called UT1 u and 27 seconds have been added since then.

There is a current proposal to change the way the adjustments are made such that they will be less frequent and and that proposal is being considered at the moment and and u uh so that the difference between uh UTC and UT1 will be allowed to increase uh perhaps to an hour and and uh under those circumstances there would not be an adjustment for uh for several centuries or not be an adjustment needed for several centuries.

Uh but uh so that's that's UTC. >> What was the motivation originally for creating it and kind of what existed beforehand? >> Well, the the the initial version of UTC had had two two motivations. The first one was that uh uh a time scale based on atomic clocks was much more stable than a time scale based on astronomy which was all the previous definitions had some kind of astronomical definition. Um and and the the the irregularities in the motion of the in the orbit of the earth and the motion of the earth meant that the time scale astronomical time scales were irregular. they had they had fluctuations which were hard to calculate and hard to predict and and uh so UTC was intended to fix that problem namely to have a scale that was much more uniform.

Uh at the same time it was recognized that uh that there were a whole bunch of users of UTC who were who had various astronomical applications uh including satellite orbits and and various astronomical things. So that it was important to maintain a link between UTC time, atomic time and astronomical time. >> And that was what the leapsecond business was supposed to do. So it was supposed to be a system that was sort of u universal.

That is it provided it provided the stability of atomic time with a link to astronomy through the leapsec. Now the the leap-second business has turned out to be quite cumbersome which is why there's the idea of changing of decreasing the interval between leap seconds. So that the the difference is allowed to increase to an hour. Uh and that's mostly because the the uh the applications that depend on atomic time or atomic time interval have become much more important than the atomic than the applications based on astronomy >> and satellites.

In the past, they wouldn't have an atomic clock on board, but now they do. >> Well, yeah. I mean, GPS satellites have an atomic clock. >> Exactly. So, they don't need to be used or anything like that. >> That's not uh >> not every satellite has an >> Not every satellite has it. U the the GPS satellites have active atomic clocks, >> but there are a number of satellites that are are are passive. That is they they are the time in the that they broadcast is uploaded from the ground.

M >> but but the important aspect is not so much the time of the satellite as the orbit of the satellites. That is the orbit of the satellites position with respect to to the fixed stars requires an understanding of of of the rotation of the earth with respect to the fixed stars >> and that's where the astronomical time becomes important because the astronomical time is basically a shorthand for the position of the earth with respect to the to some astronomical origin like fixed stars.

So we have heard like from some previous conversations that uh the need for time synchronization it started from development of railways and trains. >> Oh yeah that that that was a rather >> that was a rather crude need for synchronization. U the the that was that was more that um that was really the origin of the time zone system. uh where where uh before the time zone system there were uh each each local area kept its own time which was based on astronomical observations of that that at that point and >> sunrise sunset. >> Yeah.

Yeah. Sunrise. Sunset, you know, and so on. And and and that's that's very difficult when you're running a longdistance railroad. You'd like to have a timetable that had was more predictable, you know, for Okay. So that so the railroads were responsible for for the the time zones or the the the idea of time zones where there was a fixed time which varied by by by a simple number an hour from region to region. The the uh the other the other driving term was that uh time and frequency are really related quantities.

That is time is really a counting of frequency. And at the beginning of the 20th century, uh, frequency applications of frequency became very important. Uh, mostly originally because of u radio broadcasts that is radio radio stations had assigned frequencies and there was a need to for them to stay on assigned frequencies and so applications that depended on frequency became became important and and uh that's that that importance has just increased.

I mean the communications, power, power power distribution and so on. All of these things require more and more stable um either either frequency or time interval uh and and that those those requirements exceeded the ability of the astronomical time scales. There was just too much irregularity. >> Was the original customer the military or like the cold war kind of like military? >> No, no, no. This way before way before way before way before way before the original customers were were were commercial radio stations >> to keep them on frequency.

I mean this was u the the the uh >> they needed like an exact frequency >> size channel. Yeah. >> Well I mean for example if you're if you're if you're a radio station you you have an assigned frequency and and you don't want to interfere with other people's stations. So you have to be on you have to be on the right frequency. So, so there has to be a standard of frequency that all of the all radio stations can follow. >> Uh, and that required a standard of frequency and not only a standard of frequency but a standard of way of distributing the frequency. >> I think that there's not a this is probably obvious to you, but I think to listeners it wouldn't be obvious.

What's the connection between frequency like precise frequency and precise time? >> Oh, sure, sure, sure. They're just different different aspects of the same thing. I mean for example if you go to the music department they will talk about a frequency that they call a above middle C and in the physics department we would call it 440 cycles a second. >> Oh for sure in terms of like the number of beats per second etc. But or the number of you know crests and and troughs but what I mean is um if you're able to provide a very precise frequency let's say to a radio station have you also built a very precise clock? >> Sure.

Like are you able to measure time precisely? >> Transform it's the opposite way. Well, but what's how do you practically actually you have to interfere or you if if you if you define the precise frequency then then you define a clock by counting those cycles. >> But so what's the what's like the let's say if you're in the 30s or in the 20s and you you're selling a radio stations like a frequency source. What's that source look like?

And then how do you convert that into a clock? >> Okay. In the in the in the 20s and 30s, the the reference frequencies were quartz oscillators. >> Okay? So, you have a quartz block that's oscillating under a voltage. You apply voltage uh and it >> there's the there's a physical vibration in the quartz >> um and and the physical vibration in the quartz is generates an electrical voltage. The electrical voltage is amplified and that's the frequency.

Mhm. >> Uh now now that that frequency is compared against uh in those days the National Bureau of Standards where there was a standard for frequency. >> So that that that that frequency was compared against the standard of frequency at the National Bureau of Standards and everybody used the National Bureau of Standards. >> So there was like a quartz crystal at NIST that was like the quartz crystal. >> Well, it was it was in fact it was much fancier than that.

But that's the idea. There was a standard of frequency at NIST. Okay. >> It was much fancier than just that. But that's that's the idea. there was a standard of frequency and and and that standard of frequency was compared against the local in-house frequencies at all these places and now those those in-house frequencies were used to control the stations. Now, now NBS, which was what in the time it was not NIST, it was NBS compared its frequency against the frequency of other laboratories and that was done with radios transmissions.

That is NIST would transmit its time its frequency on WWV on the radio station WWV and that would be received in in other countries and that would allow them to compare the NIS frequency with the frequency of other countries. So there was an international community of of stabilizing the frequency. >> So would it be better to say like maybe not the time but like sampling so you you as precise as possible? >> Yeah. So I think the question for somebody that doesn't build clocks, right?

The question is once you have this precise frequency source and that's in the >> then you count it. >> Yeah. That's in the voltage domain, right? >> So you count it. >> How do you count it? How do you count? >> Well, what's the what's is it like in the megahertz and then how do you count that? Well, well, I think I think the answer is the the the frequency that you're counting depends on on what kind of device it is.

For example, if it's a pendulum clock, >> that's what I was thinking. Like in a pendulum, you can just connect it to an arrow and it goes like this. >> Well, no, no, but the the pendulum clock has a thing called an escapement. >> It's the the gear that clicks every time. >> And and the gear clicks every time the pendulum swings. >> Exactly. >> So the so the counting is done mechanically. >> Yep. Yep. That is whenever the pendulum swings the escapement advances by one by one click and that those clicks are used to drive the hands. >> Okay, that's right.

They used to drive the hands. Now effects in in a quartz oscillator the electrical voltage u you is is used to produce pulses and the pulses drive a second hand. Uh but the concept is the same. There's there's a there's a stepping motor that that if you look at a coarse oscillator it goes That is every time the coarse oscillator counts a certain number of cycles, it sends a pulse to the second hand, advances it by 1 second. >> So there's something that slows down like if the quartz oscillator is in a mega megahertz range or something. >> Well, in in a wristwatch, it's not in the mehz range, it's in the kilhertz range, but the principle is exactly the same. >> So there's something that slows it down where there's a down conversion from kilhertz to hertz >> counts.

It it's not it's not a down conversion, it's a counting. How do you how do you turn 1,00 beats from the quartz crystal into a single secondhand movement? >> Okay. Well, the the answer is do you know what a do you know what a flip-flop is? >> Sure. >> A flip-flop is something that that oscillates between two states. >> So I send in a pulse and the and the the system goes into one state and I send in another pulse the system goes back to another state.

Then I take that and that the net effect of that is that that has divided the input frequency by a factor of two. >> Mhm. >> Because it takes two pulses for the flip-flop to go from the lower state to the upper state and back to the lower state again. >> Okay. So that every time it goes to the lower state, it happens every other pulse. I've divided the frequency by a factor of two. Well, now I take that output and connect it to another one. >> Mhm.

And now I've divided the frequency by a factor of four and a factor of eight and a factor of 16 and so on. And so the frequency keeps getting lower and lower. And if I choose the initial frequency to be exactly a power of two, then I wind up with a exactly one pulse per second. >> So in a in a in a wristwatch in a in a in a cheap wristwatch, the frequency is exactly the input frequency is exactly a power of two. M >> that is the quartz >> the quartz oscillator is vibrating at 32,768 >> binary it's like two to the power of n of >> well it's two exactly two to a power and I divided by two divided by two divided by two divided by two until I wind up with one pulse per second and that drives the second hand.

Yeah. And that's amplified and actually moves that. >> Yeah. I mean it it well but but but but but you have to you have to distinguish between between engineering and fundamentals. >> Okay. The amplifier is engineering. >> Sure. Sure. >> But but the but the divide by two is a fundamental. That's fundamentally what's going on. >> Y >> and and now the only difference is that in a cesium standard the frequency is much higher. >> In the cesium standard the the frequency is 9 gigahertz.

It's in the microwave region. So I have to do many more divisions. >> Okay. But it's the same idea. I divide it, divide it, divide it, divide it, divide it, divide until I wind up with a frequency that I can drive a a display, a counter. >> I think it's also a good question like how to how to precisely measure frequency. >> Well, what you what you the way the the normally the way you precisely measure frequency is by measuring time interval. that is uh let's suppose you have a frequency which you claim to be uh a megahertz 1 million cycles per second. >> So you you count 1 million cycles of that frequency and you now compare it against the standard >> but you need to so this feels like it has some kind of almost circular argument inside because you need to have the standard in order to do this.

How do you generate the standard? the the standard I generate the standard by starting out from quantum mechanical principles. >> Okay. So you don't have to look at you're not like looking at the stars to make no no stars. No stars. I I assert that if I have >> based on the energy levels. >> Yeah. Yeah. If I have if I have a cesium atom and and I have an oscillator which induces a transition in those in the cesium atoms then that oscillator by definition is running at a known frequency.

But but there's of course like a gray there's a gray zone around that because it's you know the atom could be moving and >> effecting. So so what what now you what you've now strayed from physics into engineering. >> The answer is >> if you had an atom at perfect atom at absolute zero in a perfect space. You you know the frequency because you know the energy >> and now and now what you've discovered is that an engineered clock is much more difficult than an principal clock.

Okay. Engineering clock must in fact deal with all the issues that you've raised >> that the atoms are moving, the atoms are jiggling, the line is is is not infinitely sharp. It's broad and and now and now you've entered quantum engineering. >> And that's why cesium standards cost $70,000. But but but but you must when you stray from the principle, you get bogged down on all these engineering details which are which are perfectly valid if you were actually building the clock.

But when you talk about the clock, you don't have to talk about that. You you right you talk about quantum mechanics. >> So there's this kind of like bootstrapping challenge which is you need to have a standard in order to like improve on it or whatever. So you have to somehow start with a standard. >> And the answer is initially in 1955 the standard was astronomy. >> The first cesium clock the the t equals0 cesium clock was compared against astronomy. >> Okay.

So you had to refine it until at least it was as good as the >> Okay. So, so what I did was I ran a cesium clock for three years from 1955 to 1958 and I counted the cycles and I compared it against the astron an astronomical time interval and I used that to calibrate the first cium clock. >> And just to be clear about the working principle of the cesium clock is you have a cloud of cesium atoms in a vapor and you >> and a beam >> in a beam.

Yeah. And >> so can you explain exactly like what's the working principle of the system? >> Well, okay. So, so so, so but but you have to remember that now we're entering engineering. >> But we like engineering. >> No, no, no. Fair, fair enough. But, but separate the different ones, >> but you have to distinguish between between physics and engineering. >> Well, let's say okay, let's say we're just looking at the physics.

If you have a single cium atom in a platonic universe, you know, and you shine a laser on it, you okay? This is the question. What? Okay. So let's talk about the principle. >> Yeah. Assume there's one atom. What are you doing? >> I have I have one atom and it's in a known state. >> Yep. >> Okay. That I have somehow prepared. >> Let's say you pro you're just uh applying a field that's >> No, no, no, no, no. You remember we're talking about physics now. >> It just converges to it ground state. >> It's in the ground state or >> No, it's more than the ground state.

It's in the ground state where the the the spin of the nucleus and the spin of the outermost electron are opposite. >> Okay. Okay, they're like this. >> Okay. >> Why is it required? >> Sorry. >> Why is specifically >> Because I said so. And this >> because I said so. >> Okay. The answer is I have to choose a system. I see. That's the system I've chosen. >> Yeah. >> Okay. I could have chosen some other system, but that's the system I've chosen.

I've chosen this system. >> Okay. >> Okay. Where the spin of the nucleus and the spin of the outermost electron are like this. >> Okay. And now I shine a electromagnetic signal on that and I adjust electromagnetic signal until the state changes to this. >> Okay. >> Okay. All right. Now that's the physics. That's not the engineering. That's the physics. At that point I announce what the frequency is >> of the signal that you use to change the state.

Okay. This >> that frequency is 9192 631770 cycles exactly by definition. Okay. So, so that's an international definition that when I induce that transition, the frequency is 9192 631770 cycles. Okay. Now, that was decided by running a cesium clock against an astronomical observation. >> Mhm. >> Okay. that is I locked I had an I had an oscillator that was locked to the cesium transition and I counted its cycles and I counted its cycles for three years. >> Mhm. >> Right.

Well, that's I mean remember this is we're not talking about engineering. We're talking about concept and concept I counted those cycles for 3 years and I compared it against an astronomical interval and I asked how many cycles fit into that astronomical interval >> like a calendar basically like exact time how do you also fix the uh the astronomical because it may also shift but wait you want you want to go into engineering >> physics physics >> do you want to go into physics or you want to go into engineering >> sure let's start from physics >> let's say there's some sort of astronom of A calendar. >> A calendar. >> No.

No. No. It's not a calendar. It's an event. >> Event. >> Okay. For example, I watch when a certain star passes right next to the moon. >> Okay. >> Okay. And I wait for three years until that same thing happens again. >> Mhm. >> And I count how many cycles of the cesium beam have elapsed in that time interval. >> I see. >> Okay. And I And now that number of cycles, that was three years measured astronomically. And now I I I have three years measured in cycles and I take those two and that defines a CCM frequency. >> So you still So if in the absence of the astronomical measurement is there another way to calibrate the clock or you really need the astronomical measurement? >> No, because I needed because because the concept of of a new definition of frequency is I want to maintain continuity with the old definition of frequency.

So I want the atomic time second to look just the same as the astronomical second you or or not the atomic time second the atomic time interval to look exactly the same as the astronomical interval. >> Right? >> And the way I do that is by adjusting the number of cycles so that the number of cycles in in a cesium second matches the number of cycles in astronomical second. And the result of that is that nobody can see that anything has happened.

When I start counting cycles in cesium, it looks just like astronomy. >> I have a question on wait before you but you understand the idea. >> Yeah. >> Yeah. Yeah. >> Okay. All right. And and and now and now when you leave, >> but that means in a way that you're not measuring time absolutely. It's more like uh there's some >> What do you mean? What do you mean by measuring time? Absolutely. Well, so rather than absolutely would be for example like you have the atom in a platonic world is just sitting in the you know just one atom in the universe right >> you look at the frequency you know and you measure it and you say okay this frequency based on the energy which is somehow determined some other way >> I don't know it well enough >> I see so there's no way to know the energy >> well I no it's not that I know >> can be calculated from just the equations like >> sure but I don't know the constants that go into the equation well enough >> okay Okay, >> the way I described it is much more accurate. >> Okay, there's nothing wrong with your way in principle.

I could use the mass of the electron and the charge of the electron and go into quantum mechanics and rank along and so on and so on. >> But those constants were measured using astronomical time at some point >> effectively. >> Okay. >> Well, you take what is what is called I guess a second >> like from ast a astronomy let's say. I mean we can come back to maybe the history of the second and then you say okay second is equal to that number of ticks on the >> okay but but but it's not a second it's a day in fact you don't you use a much longer time interval >> okay and so so >> but you can always convert back to >> yeah but so so the way it was done was to maintain continuity with the astronomical time scale >> okay because I wanted that when I switched to atomic time nobody could tell the difference >> right u but historically speaking about the second uh what's like this the the the history of uh what what has been known to be a time equal to like a second and >> so so back in in antiquity there was there was no such there was the the the what was really the fundamental time interval was the day >> the the the the and and depending on on how you want to measure the A for example, you could look at uh twilight.

The day starts at twilight and runs to the next twilight. Now, that's not a well- definfined quantity question. What do you mean by twilight? Okay. Or you could talk about a day being when the sun is exactly at the meridian. Well, that's hard thing because the sun is a big gloppy thing and it's very bright and it's hard to measure. Okay? Or you could talk about a cidurial day when you look with a telescope and you ask when as the earth rotates around when some star is exactly overhead and I measure that the next day.

Time interval was based on a on some repetitive astronomical observation like the passage of the the meridian transit of the sun, meridian transit of a star u uh the rising of the equinox. I mean there there was some astronomical effect right now that astronomical effect was subdivided. So that that that interval between between noons or midnights or whatever you want to call it was subdivided into minutes and seconds.

But that was an interpolation on the astronomical on the astronomical time interval. >> I see. Well, where did it come to? Why is it not something like 10 like decimals and like all the other measurements was done like in terms of length or like let's say why is it why does it come like 12 60 >> oh that that the Babylonians like 12 and 60 but those are that's perfectly arbitrary could be anything >> I see >> I mean >> was there any like a myth or story behind why did they select >> 12 and 60 are very good numbers because they have lots of factors so that that it is very natural to choose a number like 12 or 60 because it's evenly divisible by a large number of other by a large number of of factors.

But there's nothing fundamental about 12 and 60 or 24 hours. M >> those are I mean originally the idea was I was going to subdivide the an astronomical time interval into subunits that were that were convenient and and 12 and 60 turn out to have factors to be convenient but but but there's nothing fundamental about them. Maybe I don't maybe I don't understand the question. >> Um well I I understand that it's something related like we'll say to the cycle or to the circle.

Um maybe then they started to like divide the circle into like quarters like >> but but the but but the the 360° in a circle came from the same kind of reasoning. >> It was it was that those numbers have a large number of divisor that are even. Okay. There's there's and and so you can make fractions of those numbers without without any roundoffs without any remainders. Uh but but there's nothing there's nothing fundamental about that. >> Okay. >> Like if I would do it, I would I would like say like cutting the cake.

I would do half quarter then it it would be like a binary like you know as a side. >> I think that that's fine. But if you do it that way you don't wind up with 60. You wind up with 64. >> Yes. Exactly. >> So you you you you have discovered a different system. and and and if you are in the computer business, you're using binary and you're probably using 64s, but but but that's that may seem more natural to you, but that didn't seem more natural to the Babylonians. >> I see. >> They thought of factors of of of 12.

Well, there's two and three and four and six. Okay. And and and the factor of 12 has many more factors than factors of two because a factor of two is only a factor of two. I mean, that's it. Okay? Whereas 12 has two and three and four and six >> and 60 has in addition to that five and 10 and 15 and 20. Okay. So they're all you you can make you can make divisions very easily without dealing with fractions. >> So I would say that the 60 system is better than the decimals than the binary system. >> I mean at some point I guess we're heading towards like the the appearance of internet.

But you see now you understand that you have now >> leaped forward. >> Yeah. So jump about 70 years. >> Yeah. >> Okay. So so so um so let me let me let me go. Okay. It's important to understand the principles before you talk about the engineering. Okay. >> Sure. >> And so so so let's talk about the principles. Let's suppose I have a clock and I want to tell you what time it is and and you are you are some distance away from me.

So we're not okay then I'm going to send a signal to you which says it's now 2:00 or now sometime and you are going to set your watch but my I'm going to encode on that signal what time it was when I sent it. Okay. I'm going to say on my watch it was 2 o'clock, but when you receive it, you don't receive it at 2 o'clock. You receive it by some time later. >> Mhm. >> Which has to do with a transit time. That is how long did it take the signal to get from from me to you. >> And there has to be a way of knowing what that transit time is.

Okay? And because in order for you to in order for you to in order for me to transmit information to you, you have to know the channel delay. That is how long did it take to get the signal to go from from me to you. Okay. Now that that that is a fundamental question that is not we haven't we haven't entered engineering yet, right? We're we're still back at the fundamental question that that in order for me to transmit time, I need to measure the transit delay.

Okay. Now or I need to determine the transit delay some way or other. Uh and somehow I need to communicate the transit delay to you. That is we have to agree somehow on what the transit delay is so that you will you will adjust your clock so that that you take into account the fact that I said it was 2 o'clock but you'll add the transit delay because by the time it got to you it was later than 2:00. And now I'm not going to deal with the possibility that you are moving because if you are moving it becomes even more complicated because when you are moving the transit delay is going to be changing right because you are moving away from me or you moving towards me and so so I now have to determine the transit delay which is dynamic >> and so we're not going to talk about that that's that's just too complicated.

Okay. I mean the answer is I have to do that but I don't want to do that beginning. So the answer is how do I measure the transit delay? Well, there are there are sort of three basic ways that I can use for measuring the transit delay. The first way of measuring transit delay is I can send a signal to you and you echo it back to me. Okay? So when you receive the signal, you you echo it back to me. And I say I measured the roundtrip delay and the one-way delay was one half of that.

So, I'm going to tell you, attention. I've measured the roundtrip delay. The one-way delay was half of it. And you're going to use my my knowledge of the one-way delay to adjust your clock. What's the problem with that? The problem with that is the assumption that the inbound and outbound delays were the same. All right? That the one-way delay is 1/ half of the roundtrip value. All right? Well, maybe that's true and maybe that's not true.

In a simple medium that probably is true, but in a complicated medium that probably is not true. Okay. So that the two-way method the the the weakness of the two-way method depends on the fact that the symmetry may not be all that great. >> What's a complicated medium? >> Oh, sorry. >> What would you define as like what's a complicated medium? >> A complicated medium is anything other than like a fiber is a complicated medium. >> Okay.

Or the internet is a complicated medium. >> What makes a fiber complicated? the the what makes what makes a fiber complicated the the the thing that in practice makes a fiber complicated is that it's its channel is shared with other with other users >> that is it is rare that I get to use the fiber all by myself >> like a dedicated fiber >> if I had a dedicated fiber then the delay would be really really quite stable but if I go if I use a if I use a public fiber I share that with 10 gazillion other with a 10 other conversations and those conversations have to go through switching centers. >> That is the connection between me to you does not use a single strand of fiber.

It uses a hops that goes from me to some station and then and then from some station it goes out to another fiber. it goes to your station and it and it gets rerouted at every one of these every one of these stations because in general there is no direct link between me and you on on a public fiber especially unless I mean it might happen but it's very rare to happen. It's much more likely that I'm going to go through a list of hops.

And each one of these hops, I have to demultiplex the signal, right? I have to take I have to take my my conversation out of this channel and put it into another channel and I have to deal with the fact that all these channels are being used by many different people. So the channel can be asymmetric. The the thing can be asymmetric for other reasons. uh it could be for example that that there are engineering reasons why the signal outbound and the signal inbound are different.

Okay, for example uh let's take let's take an example where where I send the signal to you as a color of light. All right, so I'm going to send you a red signal and and and you're going to I'm going to send you a red pulse and remember you have to echo it back. Now, you can't echo back a red pulse, or at least it's not not good enough for you to echo back a red pulse. Because if you echo back a red pulse, I may not be able to distinguish between your echo and the reflection of my signal from something else. >> Mhm. >> Because my signal hit something along the way and a little bit bounced back.

And the result of that is that you're going to send you're going to send a signal back as a blue signal. So, I'm going to send you a red signal and you're going to send back a blue signal. And now I can distinguish between the signal that I sent to you and the signal that comes back. Because if I see a little bit of red flash, I know that's not you. That's some reflector along the way. Okay? Whereas if when I see a blue signal, I know that came from you.

Well, that sounds like a classic idea. But the trouble with that is that red and blue travel at different speeds. Okay? Or now the effect in air is very small, but the effect on fiber is not small. Okay. So that and that's called chromatic dispersion where where the color the speed depends on the color. So the inbound and outbound are not equal. All right. So so you understand that the two-way method sounds delicious except when you get down to the engineering you discover that there are difficulties with the with the engineering.

So I say okay I didn't want the two-way method. Okay. Or well the two-way method has its problems. I'm going to use a clever method. What I'm going to do is I'm going to send a signal and I'm not going to send it to you. I'm going to send it to both of you. Okay? And and what I'm going to arrange is so that the the delay between me and you and the delay between me and him are the same. Okay? So I send a signal at 2:00.

It gets to you at 2011. It gets to him at 201. Okay. You compare clocks. >> Mhm. >> Okay. and you decide that your clocks say the same thing so that you can compare the clocks without knowing what the delay was because the delay to use the technical term is common mode. It adds to both to both numbers as a subtractor of the difference. >> But then we have to be the same distance from you >> or pretty close as close to the same distance as I can arrange it >> or or the difference in distance has to be small, >> right?

Deminimus. >> Okay. So that so that's called common view. Okay. So so so that so that method for example you hear that all the time. If you if you go out to if you go if you sit in the yard here, you'll hear the carolon from the state from the from the arts and science building. Okay. So, the carolon transmits a a ding at 3:00 and everybody receives the ding at 3:00. >> Mhm. >> And it doesn't matter if the carolon is a second late because everybody says it is now exactly the same time.

So, everybody synchronizes synchronizes their clock to the same time even if the carolon is not exactly at the right time. and without knowing what the what the transit delay is provided the transit delay is about equal. Okay, so that's what that's why the carolon that's why the carolon is a carolon. Okay, it transmits time in common mode. So everybody sets their watch and everybody sets their watch at the same time.

Okay, so that's so that's the second method. So so the first method was was two-way. It depended on the equality of inbound and outbound >> peer-to-peer >> peer to okay the second method well I I would say actually it's not peer-to-peer it's it's more uh it's more uh master the follower okay that is uh it's really a more master to follower the common common view is really peer-to-peer because in common view I send the signal to you too and I don't participate in the difference you participate in the comparison >> so so so so common view is really closer to peer-to-peer Okay, the the the two-way is really closer to to to to lead a follower.

Okay, I send you the time and you follow it. Okay, whereas in common mode in common view, you you you compare clocks together and the accuracy of the comparison does not depend so much on my accuracy because it cancels in the difference. Okay, so that's method number two. Okay, method number three. Method number three is u I know that the delay between you and me varies with something else like temperature. >> Mhm. >> Okay.

So for example if I calculate the speed of sound and air just as an example it depends on temperature. So I measure the I measure the temperature and I say aha I know the speed of sound because I can measure the temperature. So I tell you you measure the temperature I measure the temperature. We measure we estimate the temperature along the path and we calculate the speed of sound from knowing the variation with temperature.

So those are three methods that are used for synchronizing clocks. Okay. So you understand that that that that the whole business with synchronizing clocks depends on how you got to determine the delay. >> Mhm. >> Okay. That's the root of the problem. So the root of the problem in this in a clock synchronization is measuring the delay. Okay. And now where the delay varies I have to sort of I have to play a tricks for measuring the delay and and and the internet >> going to engineering. >> Yeah.

It it but it but it's but it's it's not even it's not even generic engineering. It's very specific engineering. I mean the common view idea is kind of generic engineering right but but but the internet time service depends on specific engineering. namely I have to actually calibrate the exact link that you you use that that we use. Okay. And so I perform all sorts of tricks by sending signals back and forth and I try and estimate what what the delay is not in a generic sense but between your between my channel and between the channel that links us.

Okay. And and to the extent that I can do that I could transmit the time accurately. Okay. So the the entire process, if you look at all the papers, the entire process is a question of how I detect the how I measure the delay. All right? And it's hard. It's hard. Now, the GPS system uses a combination of all of these methods. Okay? The GPS system broadcasts the properties of the atmosphere or the properties of the ionosphere.

So you can calculate the delay. It it calculates the it uses the common view method. So you receive the signal from one transmitter in two places. Okay. And and it doesn't use two-way, but the two-way method exists. The two-way method is used with passive satellites where you send a signal from here up to a satellite and the satellite echoes it down and then the other signal goes back up the other way. And that that's a standard method and that depends on the symmetry of the of the of the path in two directions.

Okay. I don't know if I've answered the question or if the question has gotten lost. >> Yeah. Yeah. No, it's it's it's indeed a >> but but you you understand that that in the end that's really the hard part. Okay. That in fact when you when you see my clock, you really don't see my clock. You see my clock through the channel that links us and and the the inadequacies of the channel are the dominant problem. If the channel were perfect, the job would be easy.

So, for example, if we have a piece of wire, >> if we're not too far apart and we have a piece of wire, the job is easy. >> Is there a way to encode something in the signal that changes as it passes through the fiber such that at the end when you receive it from the signal itself, you can see how much time it's spent. >> I don't know how to do that. Uh, if you got a way to do that, speak up. I don't know how to do that.

I mean, you got to remember that everything I put on the signal is delayed by the path. >> Yeah. So, I've got a uh uh I I I I don't know how to do that. Okay. I I need >> Well, for the basic I mean, this probably wouldn't work for a lot of practical reasons, but like attenuation, right? So, the longer the fiber is, the more light is lost in the fiber. But then you have a lot of imprecision about how much light did you put in. >> Okay.

Essentially, you you are you are bringing up the amplitude stability. I mean I mean that's just a fancy way of saying what you just said that that I'm bringing up the amplitude stability and also the detection efficiency. I'm going to bring up all sorts of problems. You know, in return for solving this problem, I now bring up another problem. So, so, so you understand there's that that you're now into engineering. Okay?

And you got to ask, well, which one of these problems is easier to solve. >> And, and the answer is, well, I don't know. But I'm not sure I can give you a general answer for which problems which one of these problems is going to be easier to solve, >> right? It depends on on you know who are you and where are you and and and you know what are you and what do you have and and uh and uh and there's no there's no one perfect solution that fits all situations >> how does it work today like when uh when clocks are synchronized today across the internet how does it actually work >> well the the the answer is most of the clocks synchronized along the internet use a two-way method >> I send the signal out you send it back I divide the I divide the roundtrip path path by two and I and that I I use the one-way path and the answer is that's sort of right but not completely right.

Okay, it turns out that this the asymmetry is a few percent. >> So for example, if if I if I measure the roundtrip time to be uh I don't know 100 milliseconds, a tenth of a second then then the asymmetry is going to be a few milliseconds. In other words, the difference between the inbound and outbound delays is going to be a few milliseconds, which means that the accuracy of the time transfer is going to be on the order of a few milliseconds.

Well, is a few milliseconds good enough? Are you happy with a few milliseconds? And the answer is if you're happy with a few milliseconds, peace and hallelujah, you're done. Okay. On the other hand, if you're interested in a microcond, internet is not good enough. >> And the reason is that the asymmetries are too large. And so so it really depends on on not only how well the thing is going to work but how well you need it that is you know >> so how does the highest performance synchronization work? >> Well the highest performance synchronization is two-way through satellites >> and the reason it is the highest performance is that the paths are very symmetric.

So that the taking the roundtrip delay and dividing by two is a very accurate measure of what's going on. Aren't there some like u you know the wind changes in the air pressure those are those are negligible effects okay they're negligible effects the the now a dedicated a dedicated wire or dedicated fiber would be very good >> and the the only trouble with a dedicated wire or dedicated fiber is they're very expensive >> unless unless the distance is very short but but if the between here and and uh east coast the the fibers are very expensive a dedicated fiber would be very very expensive too expensive.

You wouldn't want to pay it. I wouldn't want to pay it or I can't pay it. >> Okay. >> How How did you originally get into the clock business? What's the story? Like where did you grow up? How did you get >> I I well I >> I I started out in atomic physics. Uh I I was trained my graduate program was in atomic physics and when I came to Jill I started working on atomic physics problems. uh and uh I was I came to Jill as a posttock in 1967 and and uh the posttos were two years that is at that time it was very strictly two years.

So, so in 1969 when I started looking for a job, I was offered a job at NBS. What was was it was then called NBS and and uh the work that I was working on at that time uh had to do with measuring the speed of light that that was there was it was a rather large program uh to measure the speed of light and that was going to be part of the redefinition of the meter. that as the intention was that that the meter would be redefined in terms of of the frequency and the speed of light and that would redefine the meter.

Um and now when that program finished uh it turned out that the group that was that was part of that was reorganized into the time of frequency division uh that was a management decision to to move that group to the time of frequency division. So, we all got involved in the time business. Uh and and uh uh we didn't start out in the time business. We kind of wound up in the time business and and uh and when you're in the time business, you you begin to think about clocks and time and frequencies and so on and so on.

But almost all the people in the time business came from something from some from some other field, usually atomic physics, usually atomic structure. Uh and and they they moved into into clocks by some by some accident often. >> And at the time was it like a hot field to get into clocks? I guess there was like the space program, right? The moonlight. >> Oh, this was before the space program >> was in progress or something. >> I don't I don't remember that it was a particularly hot field.

I I don't uh I I don't remember. I I don't uh uh I think it was more a a field that attracted uh precision measurement people uh uh that as people who are interested in in the last decimal place >> uh and and because that's what the clock kind of that's what the clock business is. It's a last decimal place kind of kind of position kind of kind of work and you have to be into the last decimal place kind of work. uh and there are people who are not there are people who who uh who would like to understand the the fundamentals of the universe but are not interested in the last decimal place of the understanding and and uh and that's a fine thing too.

I mean that's that's perfectly and so I guess that that the clock business is the place where you go if you want to be interested in in um in that kind of last decimal place sort of stuff. >> And how did you get into physics originally? Like what's the story? How did you get into science? I I I guess uh I guess I was always interested in in um uh in electricity and and uh and radio. Uh even as a child even wasn't a child, but when I was when I was when I was a kid, I was I made radios.

I made radios. Uh I built radios. I built I built electrical electrical stuff and and that kind of led natural I don't know if it led naturally but but uh but that was sort of if you wanted to get good at that that's where you went to uh although I mean there are there are other ways of doing it you could have gone into electrical engineering or whatever but but but physics was the opportunity that I had and so that's the opportunity I had >> looks like radio was the inspiration toy for a lot of uh young kids at the like that time.

I know there was like Fineman was talking about. >> Yeah, that's right. But see, but this is before all of that. But that's true. >> But but this is before all of that. >> I think the the uh the choices were were uh were physics or or or chemistry or biology. And at that time biology was not a very interesting subject. I mean this was before DNA, before the double helix, before all that stuff. And uh I mean biology really changed but it wasn't interesting at that time and and uh and so physics was sort of the uh I guess that was I don't remember but it was sort of a natural thing to it was sort of a natural thing to do if you were interested in playing around with with with with stuff you naturally gravitated towards the physics department.

And this was you you were originally from New York or or >> Oh yeah, I grew up in New York. I I I was uh I I was in New York for uh through the undergraduate school. I was in New York till I was 25. >> What was New York like back then? >> Uh well, I lived in the Bronx. Uh and and uh uh the Bronx is the kind of place where uh where tourists don't go. It's not a tourist place. Uh and so uh I lived in a in a in a lower middle-ass neighborhood.

Uh uh my father was a public school teacher. He taught the sixth grade. Uh and uh he taught sixth grade in the public in the in this in the public the city public schools. Uh and in those days and even now uh sixth grade teachers were not well paid. Uh and and it's probably still true. I don't know where they get paid now. They probably get paid more now, but but six or eight teachers didn't get paid a lot. So, so u uh we lived in a we lived, you know, we lived beyond the end of the subway, okay? >> Which meant that it was a it was a hassle to get anywhere because because you had to take a bus to the subway.

It was it was a real hassle. >> And how robust is the system? So let's say robust like for example if there are like hackers who would try to to break it down or like try to synchronize all the >> I think I think there there are whenever you have a system there are going to be hackers and and um and yeah it's a problem okay >> how do you protect it? >> Uh well the answer is there are two ways of protecting two general styles of protecting.

The first style of protecting is to have many different sources. So for example, the N system has 25 25 servers. So in order for you to to attack the NIS system, you have to attack all 25 servers >> or like 51% of them type of thing. >> Well, whatever. I mean, you name a number, but but the point is attacking one server is not enough. Okay? uh that I can I can ask for time for many different servers and see if they see if one of them is different.

Okay, so that's that's method number one. And method number two is eternal vigilance. You watch. Okay, and you have checks and you have you have you look you watch and and uh what can I say? You watch. How would you what is like if you were building like a super weapon to desynchronize the clocks of you know we have a theoretical uh nemesis and they have synchronized clocks how would you desynchronize their clocks with like a >> well what first before before we get involved what accuracy do you need do you need milliseconds do you need microsconds or do you need ncond >> that's a good question like what what is the accuracy threshold where things start to fall apart that are important >> well if you're talking about public internet then Public internet is a millisecond level service. >> Okay.

Uh that is because there are all sorts of asymmetry problems and noise problems and so on and so on and and you're going to get millisecond level service but not not dramatically better than a millisecond. So if you need if you need millisecond level service, you're home free. You're done. If you need better than a millisecond, it's not so easy. Now if you if you uh if you have a good quality if you have a cheap GPS receiver you can get um a microcond maybe a tenth of a microcond with a cheap with a cheap GPS receiver you know something that you find in a in a cell phone >> okay you probably get a microcond without without any without any real problem uh and and and on the other hand the display of a cell phone is probably not good to a microcond But that's not the receiver, that's the display.

They simply they simply figure you don't really care about the time to a microcond. When you get below a microcond, it begins to get hard. Okay? Because the delay and the fluctuations of delay and the knowledge of the delay become hard. When you get down to nanoseconds, it becomes really hard. You need amateurs can't do nancond. Even professionals have difficulty doing nanocond. the the timing laboratories are at the nancond level or maybe maybe two maybe three. >> Any technologies that rely on nancond? uh the main technology to rely on on nanoseconds is not is the equivalent in in the frequency domain that is it's stabilities of a part 10 to the 13 or part 10 to 14 u and and there there the frequencies are in the telecommunications in in in 5G networks and cell phone networks and so on uh and so tele and navigation because uh the speed of light is a foot per ncond roughly speaking and that means means that if you want an accuracy of a foot, you got to have an accuracy a timing accuracy of an >> nancond. >> So if if you're on the other hand, a cheap GPS receiver is not a one foot kind of kind of instrument. >> Uh it's a 50 foot kind of maybe a 25 foot, you know, you know, turn right at the next traffic light.

Uh well, you don't have to get it down to a foot. You get it down to 20 feet maybe. >> Uh on the other hand, you know, uh send a bomb through that guy's window. Um well the accuracy is probably a foot and and so you need you need timing accuracies on the order of a nanoc. >> So so navigation serious navig on the other hand uh and after all an airplane is a big thing and so you don't really care about where an airplane is to a foot you might care about where an airplane is to 100 feet okay or 20 feet or 50 feet or something some number.

So it's so it's tens of nanconds and and uh and that's a GPS level. So so uh or radar I mean I think ground control ground control is probably radar but but uh GPS is long to long distance navigation is probably done with GPS >> and yeah how would you how would you build like a super weapon to disrupt the uh disrupt clock synchronization? Well, you got to remember that that that the clock standards are distributed. There's no single point of failure or at least that's the intent is there's no single point of failure.

And so, it's designed exactly for that reason that that that uh that uh there's no one clock that that that maintains the only standard of of time for the whole universe. There's a whole ensemble of clocks. They're located different places and and and so so I'm not sure a super weapon is is uh I mean it it it's a distributed problem. Now now for example you could say well okay uh I'm going to destroy the whole GPS system. >> Mhm. >> Well the GPS system has got 30 satellites. >> You have to destroy you know it's a big job.

It's not it's not it's not >> But for example, couldn't you create a false clock? Like instead of having 30 GPS satellites that you destroy, you just put up 30 more putting out. And that's called confusing confusing information. >> That's called spoofing. And spoofing is generating signals that are false but appear to be true. >> And there's all there's a whole industry in spoofing. Okay. Especially in military theaters.

What about just warming up? Let's say you know where the satellite is and you just warm up the cesium clock in the satellite with >> I like you that's hard to do because I built a cesium clock and I'm not going to let you I'm not going to let you get your hands on it. It's much more it's much easier to mess around with the signal on the ground. >> Right. >> So the spoofing spoofing guys tend to mess around with the signal on the ground >> and and and and there are there are jamming and spoofing attacks >> that's common.

Well, I mean, you read about in the newspaper every day. The answer is, yeah, the military is engaged in a continuous uh >> But if the if the signal is encrypted, how can you spoof it? Like, is it not encrypted? >> Uh, the answer is like everything else, it's a race between the offense and the defense. Okay? And the if it's encrypted, then I break the encryption or I understand the encryption or I decode the encryption.

And and and now if you think it's a perfect encryption, then you're making a mistake. It's not nothing is ever perfect, right? But let's say if I shine a laser at GPS satellite and I'm, you know, I find the cesium >> and you burn it up. >> No, you just warm it up slightly. >> Well, I just to make the time >> I think I can shield I think I can shield myself against that >> just just by building the satellite with enough. >> Yeah, it's just a mechanical problem.

You you got to be clever than that. >> But but but that doesn't mean Okay, you you you send off a bomb and blow up the satellite. I mean, it >> that's too that's too much. I'm just thinking of like what are subtle are there any subtle ways to like >> but the answer mess with the >> but the real answer is that I don't build the satellites that I don't know >> okay but the people who build satellites worry about those kind of >> like solar flares change >> solar flares solar flares can be disruptive because they change the delay >> because the electron density in a solar flare can affect the delay and it can affect the calculations of the >> so could I fly another satellite next to the GPS sat satellite and just release ions that change the delay next to the time like the flight the flight path of the photons. >> I think now >> could I just put out something that has like a high refractive index >> in the flight path of the photons? >> Can you define that as an act of war? >> I mean for sure I guess it would I guess it would.

I'm just curious like what are the what are the so the vulnerability is the is the assumption of the refractive index of the path path to >> the but but let's talk about in principle okay let's not talk about the details if you if you damage my satellites in a way that I can detect it I'm going to attack you >> no but I'm it's not not you don't touch the satellite so there's a between the satellite and the earth there's like a cone where the signal >> and somehow you do that without my detecting it. >> Yeah, you would have to like uh heat up or change the refractive index basically somehow of the cone. >> I think the answer is >> very difficult.

I >> I Well, I would guess it's difficult, but but but you're kind of on the edge between theory and engineering. >> Yeah. Yeah. >> Okay. And the answer is yeah. I mean, the answer is uh historically that's not the way the attacks are done. the attacks are done on the ground. >> It's not like a laser beam also. It's it's like, you know, more like a directed RF way. So it goes like everywhere. >> Yeah. But but then on the ground, but historically that's not how it's done.

It's done by it's done by jamming and spoofing by by rogue transmitters on the ground. That's in fact how it's done. Or in balloons or in airplanes. >> Uh it's not done by attack at least not yet done by attacking the satellites because because that's that's hard. >> Yeah. Yeah. and and and I don't is it hard but I'm I build a satellite and I think about all that stuff and and and if if somebody says well what about when they attack the satellite this way I build it so you can't it doesn't you know I mean uh and and and the answer is but I don't build satellites I don't know how they do it okay but but but if I were building satellites I'd worry about that right and and >> for sure >> what what about the fiber system like if you can you inject >> yes >> signals into the fiber like >> sure the question there is detect is inserting them without being detected. >> Mhm. >> Okay.

You got to be inside the phone company. >> Okay. Yeah. >> The answer is yeah. And and that is the insider attack and insider attacks are very difficult. By definition, insider attacks are difficult because an insider has access to the to the internals that that and you got to trust. Insiders are difficult. And and that's just a generic statement. I mean the the employees of NIST are are are are potentially dangerous.

Yeah. I mean >> so it's not I feel like it's it's not just like a technology. It's more like a human. It's exactly a human factors. Yeah. And and and uh employees inside the system are fundamentally a problem. But that that must be one of the ideas behind like quantum versions of this type of information. >> No, I would say that it's much simpler than that. You don't allow one person to do anything, >> right? Yeah. Yeah.

There's definitely there's there's st there's management style >> and but the answer is uh >> look all of these systems have strong points and all of these systems have if you're inside the system, you know where the weak points are. I mean it's it's >> uh Historically back as long as there have been have been have been uh uh uh these sorts of problems there there's a race between the offense and the defense you know uh you know do my do my bullets pierce the tank or does the armor of the tank too strong and the answer is yeah and when the armor tank gets stronger the bullets get stronger you know uh >> maybe yeah we we could like make maybe come back like in history because you did a lot of your uh like studied this that the research during the I guess when the cold war was in ongoing how how was it like between let's say >> oh but competing with each other and and and and the end answer is when when you devised an attack I devised it I devised it >> was there a was there a parallel Soviet effort to make like very accurate clocks at the time >> oh yeah yeah the Soviets have yeah the Soviets have clocks that are just as good as the United States well pretty electric.

Yeah, the Soviets have primary frequency standards. Yeah, they do. >> Well, at the time that you were building the cesium standard, what was happening in the Soviet Union? Do you remember? >> I don't know. I I would guess. >> Was this an area where there was open collaboration or this was considered like military? >> I don't remember. But I would guess that that was not an era of open collaboration. >> I think in some areas of science they were like pretty open, >> but I think that was I don't remember.

I I don't remember. But I don't think that was an era of open collaboration. uh and and uh uh but I I really don't remember. It's 50 years ago. I don't remember. But but I think it was not an era of of >> Were they also using cesium or do you remember do you remember? >> Oh, everybody's using cesium. Okay. >> Why why cesium? >> Well, because it's a he it's the heaviest atom that's easily available. >> If if you're going to have a if you're going to have an atom in a beam, the cesium cesium atoms are heavy and so they move more slowly.

And so so cesium is an obvious choice. Okay? There's no any physics any physics graduate student would tell you to build this clock. It's easier than >> because the line widths are smaller. >> Well, but but the line width has to the line with by the uncertainty principle. The line with depends on how long you have to observe the state. >> Oh, because it's moving slower. >> Yeah. Because it's moving slower, you can observe the state for longer.

And so, so there's a fundamental advantage to CCM. So, every graduate student knows that. I mean, that's that's sort of u and that's why CCM was chosen. It was not that's not some profound every graduate student would know that that that >> now it turns out that there were engineering issues that cesium is easier it's easy to produce cesium beams is easy to detect cesium beams but now you've entered you've entered engineering >> mhm >> but but but the the cesium is the heaviest easy atom to use.

Okay. It's the last atom that's easy to use. >> What happens after that? What what goes wrong? It be it become more difficult to detect more difficult to produce or generally become generally speaking the column one is the best the best all the atoms are all the good states are in column one the alkali atoms that's why rubidium and cium they're all cousins are all in they're all in column one and that's an engineering issue that the engineering the the atoms in column one are easier to produce and easier to detect and that's just that's just engineering it just it turns out >> is there any Morse law in atomic clock or clocks in general like let's say like the precision is >> the precision has certainly improved but I'm not sure that there's a Moore's law that characterizes it uh I don't know okay the see the the the accuracy the performance has certainly improved by by uh six or seven orders of magnitude in uh 50 60 years >> seems like it was a big jump >> well the the big clocks appeared the big there was a big jump from no cesium to cesium absolutely >> zero to one >> and then there was there was then there were incremental improvements uh and now there's another big jump to optical clocks but optical clocks are still laboratory devices like jun's clock it's laboratory device >> the the it's not atomic time doesn't run on jun devices >> were they deployed on satellites >> no atomic >> obstacle atomic clocks >> no >> can you explain what the what's the like going to higher frequencies of light why does that improve the performance or what's the what's kind of >> well there the answer is when I count more cycles in the same same time interval >> the uncertainty >> my resolution is just higher >> okay so that all other things being equal assuming everything else is equal if I make the frequency higher the accuracy gets better or the stability gets better >> and the problem the problem until very until relatively recently was that optical frequencies were too high to count >> until it get to optical combs to translate them down.

Okay. But until until the optical combs came along, there was everybody understood that optical frequencies were the right thing, but they it was simply unusable. You couldn't count them. The frequencies >> electronics couldn't keep up. >> Yeah. You can't use electronics. You had to use frequency combs. >> Frequency combs. Yeah. >> Yeah. That that's right. >> How does the what is the working principle there? uh the work of principle is that the optical frequency is divided down >> by some very large integer that you know and and and uh and so the division is that there it's a very large integer that you can measure from the comb and and and uh and that's that's the essence of the trick >> and then you can put it into normal electronics or you know high-speed electronics >> then I can put it once I get it down into the gigahertz region I can use normal electronics >> so that's the frequency com they kind opened the path towards optical atomic clocks to synchronize the yeah >> atomic transitions with the laser. >> Everybody before that everybody understood that an optical clock was the right thing to do but there was no way of counting the frequency until until the combs came along that allowed the translation of the optical frequency down into a microwave region where it could be counted.

Yeah, I think it would be great if you could also explain um that how the atomic clocks work in in the sense that like you have an atom and then you have like electromagnetic field that you are trying to >> okay well but but so let's talk about cesium beams because sure okay so so the first step in this remember you got to the first step in the in the in the business is to make a beam. >> Mhm. >> Right. You got Well, it turns out that making a beam of Cam is very easy.

You have a little a little kettle and it's called an oven, but it just it's basically just a small block with a very small hole in the front. to evaporate it >> and you evaporate the cesium and and seesium is very easy to evaporate because it it it okay you have to get it to like 50 C and it begins it begin it evaporates >> because vacuum it it starts >> it almost cesium is a solid solid but it almost it almost melts in your hand okay it melts to like 27 C uh so so that that so you heat it just gently and the stream the cesium beams stream out of the little hole in the front of the of the front of the thing so that now you have the beam of cesium so that the beam of cesium.

So that's the beginning and and so the oven is just is just a block with with a heater and a cesium thing and a small hole on the side. But now you have to set it up so that you get the atoms in the in the bottom state, right? You have to you have to you have to you have to decide remember you because you got to you got to arrange it that the atom has to be in the in the bottom state because that's how the clock's going to work.

All right? Well, the way you do that is that it turns out that the bottom state has a unique magnetic property. Okay? And so you run the beam through a magnetic field. >> You filter it. >> Okay? And it turns out that if you do make the magnetic field just right, the atoms in the state that you want get through and the other atoms are deflected out. >> So all the atoms go into the magnetic field, but only the good guys come out. >> Okay?

And the good guys are the ones in the lower state. What is the >> what's the condition for this? Is the magnetic dipole >> magnetic moment? >> Magnetic moment. Yeah. >> Yeah. >> How was this found? Like how did you guys find this filter? Like >> it was it was well >> was already known or you had to find it to >> um it was known. I mean there was a lot of at a lot of work on atomic beams for years years years. But but but there was there's some theory and then there were some measurements but but >> uh but the measurements of atomic structure date back to before before atomic clocks >> but is this something that like when you were building the clock you had to basically figure out like okay how are we going to how are we going to bas purify the >> oh no that idea >> that idea had been developed earlier.

Okay. >> Okay. The magnetic magnetic selection was known earlier. It was back in the stern experiment. The stern experiment well back you know 30 years before uh demonstrated that fact that you could separate the beam based on the magnetic moment. >> Mhm. >> Okay. Using magnetic separation. Okay. So back in the stern girl life experiment back in I don't know 1920 uh it was known that you could separate a beam uh based on the magnetic moment and and you that is and so so the states were were physically separated that is they one state was deflected relative to the other state.

So you just put a stop at the output so that only the good guys get through. >> Mhm. >> Okay. So there's a mechanical mechanical you know knife edge or step stop so that only good good guys get through. So now what comes out of the beam is only the atoms in lower state. >> Mhm. >> Right. That's all you get. Now I irdiate that that beam with my oscillator with my with my microwave oscillator and I adjust the microwave oscillator until the atoms change state.

Well, how do I know that they change state? Answer. I do the same magnetic trick again. So I have another magnetic thing except now only the atoms in the upper state pass through. So I start out with atoms in both states. They go through the first magnetic magnetic selector and now only the atoms in the lower state get through. Now I shine I shine microwave radiation on it and and when the microwave radiation is right, the atoms flip it to the upper state.

So now the atoms in the upper state and now I have another another magnetic selector and now it's adjusted so only the upper state passes through >> and then you measure a ref frequency when most of them will switch. >> How do you detect cesium atoms? What's the detector? >> Okay, it turns out that that's a magical fact of cesium atoms. >> Okay, >> and that's one of the m that's one of the reasons the alkali atoms are used in general.

If I shine if I if a if if a cesium atom hits a hot platinum wire. Okay. The cesium atom is ionized. >> Mhm. >> And and and and it comes off it goes in as an atom and comes off as an ion >> and essentially the electron is stripped off. And now I have a charge. And now I simply detect the charge. I simply count the charge >> on the platinum wire or >> no the I have another I have another cathode >> cathode off to the side and I simply attract the guys and I count them >> and that that hot wire detector is a magic of the alkalies. >> It only works essentially it only works for that's that's one of the reasons the cium was chosen. >> Wow, >> that's a really cool detail. >> Okay, so the so the the now now in subsequent years there are fancier methods.

Is there a fancy >> is the wire the anode? And so then it ends up being like a balanced system or >> well there's a wire and the cesium atoms come in and when the cesium atom hits the wire it they come out as as ions and now over here I attract them with a negative thing and I simply count the number of ions that that reach my detector. >> But the electrons do they build up on the platinum wire? >> Well they they just go I mean they don't build up they just you know they get siphoned off the ground. >> Okay.

Okay. Interesting. Um, how was that discovered? >> I don't know. But it it was it was a well-known thing long before cesium clocks. >> It was a well-known idea. It was called uh and and u uh there are there are a number of different effects that are the same effect. U there are some there are some atoms u which if they're in the excited state when they hit the wire they they send off an electron. I mean, so there there are there is a number of different a number of different configurations in in the in the the hot wire detector.

What you get off is an ion, but sometimes what you get off is the electron. So sometimes you're counting electrons and sometimes you're counting ions, but but the concept is the same. You change a neutral into a charge. >> Mhm. Was there any limitation to like you know how many charges you could detect versus how many atoms you could produce or anything like that? Like what was the signal detection efficiency was tiny?

Okay. So that that that that that that the beam is I don't know you know thousands of atoms or maybe a million atoms. It's tiny. Okay. >> So you need to amplify the the signals what you're measuring. >> But but now see now you're just dealing with electronics. >> Okay. That's the physics the physics part of it is over. Now it's just electronics. Now you just have amplifiers and detectors and meters. You know that's all.

I mean you know >> I guess you don't even need an absolute number. You just need a threshold at which you can say this is the best frequency. Like essentially if I go higher than this frequency it gets worse. >> That's right. >> If I go lower it gets better. >> Well in fact what you do is you jump back and forth across Yeah. Yes. That's right. That's exactly how you do it. >> Okay. You jump back and forth across the maximum and and you jump back and forth and and that's exactly how you do it.

You you find that peak. >> But the the width of the peak I guess does depend on like the uncertainty of the peak depends on the detector. >> Yeah. But but see but now again but you've that's right. So now the width of the peak depends on how long you get to how long you get to observe the right. >> So so now now you see you can't completely divorce the engineering but but the answer is yeah. Okay that's right. You want to make the w the peak as narrow as you can. >> And so that means you need as long a tube as you can get. >> Okay.

So so so the natural the natural thing is to make the the tube as absolutely as long as you possibly can. >> Right. Well, in commercial cesium devices, they have to mount on racks. So, the width of the length of the tube is 19 in. >> So, just to because maybe this is me being slow, but if the tube is longer, the atom travels through the tube a longer time. So, it's irradiated for a longer time. Why does that increase the accuracy? >> Because the uncertainty principle says that when I when I look at the when I look at it more when I when I when I observe a thing for more time, the uncertainty in the in the frequency is reduced.

But are you observing it for a longer time? Because the observation time is really when it hits the detector. >> No, no, no, no, no. The observation time is in the irradiation time. >> Why is that? >> Because once once the radiation once the irradiation time is over, the rest of it is just detection. The the state has already been transmitted. The the state change has already done. >> So you're saying the longer it passes through the tube, the higher the chance it has to absorb the >> absorb the frequency. >> Yeah. >> Yeah.

Yeah. Okay. There's also a question of of like what are the chances of it actually absorbing the >> Well, the answer is if it doesn't absorb it. So, but now now you're into efficiency. >> Yeah. Yeah. >> That that's simply a matter of of what's the efficiency of the absorption. >> Okay. >> Right. Do I do I manage to get all the atoms or only 10% of the atom? That's that's >> Well, I guess what it is is the atom is observing the frequency as it's passing through in a way and the frequency also has some instability.

So the longer it passes through the frequency, the more it's like averaging over the behavior of the frequency. >> But but that what you're saying is absolutely right. But but but but you've strayed into engineering again. >> Okay. >> You stayed into the question of well now how do I actually build this? Okay. But but but and so you make the want to make the tube long. You want to make the frequency stable. I mean you have to do all those things.

That's all. But that's absolutely true. No, I'm just trying to understand where this like accuracy increase comes from by making the tube longer. >> Oh, no, no, no, no, no. It's not. The accuracy does not increase. Okay. Well, actually, some say it does increase because because the the the line becomes narrower and the top of the line becomes sharper. >> Exactly. Yeah. >> Okay. But, but but but but but the accuracy has other terms. >> Okay.

The accuracy has other terms and because we haven't we haven't talked about the other terms yet. Okay? But but but but let's just stay with this. You understand this? I want to make the tube as long as I can. And so the natural length of the tube is a cesium fountain. I don't make the tube horizontal. I make the tube vertical. Right? And I shoot the atoms not horizontally. I shoot the atoms vertically. And I shoot them so slowly that they go up and they come down and they fall down again.

All right. And now the interaction region is very the interaction time is very long. >> All right. And so the result of that is that the the potential accuracy is much better. And so all of the the ultimate standards are not the devices that you see out there. They're fountains. They're not horizontal. They're vertical. >> And what's the largest uh >> fountain clock? meters. >> So they didn't try to make it >> totally back back and a long time ago they tried to make it the engineering just becomes impossible. >> I see. >> Okay.

Uh it just becomes extremely difficult. I mean there was a time uh when they they made they wanted to make a standard that was like you know two floors 16 ft high but but the engineering just becomes impossible. >> Okay. >> What's the challenge? I think the challenge is getting the beam to come down again. Oh, >> I see. It spreads a lot and then >> I I don't I don't know. I don't I've never built a clock, but >> through the walls and >> but but you see but there are there are all sorts of problems waiting to be made waiting to be talked about that we haven't talked about yet. >> Okay. >> For example, one of the problems is that the cesium atom is not completely isolated from the rest of the world.

It collides with other cesium atoms. Well, when it collides with other cesium atoms, that's very bad because that changes the energy levels. >> Molecules have to form molecules or >> Well, well, it collides with air, but it collides. Remember, this is a beam. >> So, there's a whole bunch of guys, a whole bunch of cesium atoms going up. >> Well, they collide with each other. Okay. Uh so, so that's an effect. Okay. There are effects where where you have stray magnetic fields or stray electric fields.

All these things affect the transition frequency and they they affect the transition frequency in a systematic way. That is the transition frequency when you have a stray electric field is not the transition frequency that you think you have. >> Mhm. >> How do you shield from stray fields? >> Well, the answer is you have shields. Okay. You have magnetic shields, you have electric shields. I mean, you have shields and shields and shields. >> You like coat the glass cell with some >> Oh, no.

No. uh magnetic imu metal. They're uh >> like electric fields. >> Electric fields are relative relatively easy. You just have a a a closed very good conductor like a copper a copper sheet or or a Okay. But but any good conductor is okay. >> But then don't you block it uh like a ref would be reflecting back? >> Well, you the answer is you have to have little holes. Okay. Okay. That's right. I mean, see, but but but the answer is yeah, you you make a shield, but now you have to have a little hole to to to see what's going on.

Well, that's a problem. >> I see. >> And what what was the biggest of all the noise sources and like these kinds of practical issues, which ones were the mo most like challenging and or like had interesting solutions? >> Um the collisions between atoms is is a serious issue. Uh the collisions with the walls to the extent that there are collisions with the walls a serious issue. uh stray magnetic fields. >> But how can you control for collisions between atoms?

I mean, what >> by using by using a lower density. >> So tuning it down to the point where it's almost >> That's right. But but the problem with using a lower density is you have a lower signal. >> Shot noise. >> Okay. What >> shot noise like fewer >> Well, okay. That's right. But but if you have you have shot noise but but if you have a lower density, you have a little fewer collisions. >> Yeah. >> So now the question is which you know >> like >> That's right. and and and uh but the problem with lower collision with with lower density is you have a lower you have a lower signal so you begin to get electronic noise >> or or electronic noise becomes a more serious issue >> and so why did you run the clock for three years was that an arbitrary number is there like a fundamental reason >> no because the astronomical event is hard to detect >> because you you you've got to you the the the dayto-day variation the astronomical event remember you're seeing the transit time of a star through the through the jittering atmosphere.

So there's there there's there's atmospheric cintillation and it becomes hard to detect. You've got to average over a very long period of time. >> And how did you know like when the clock when you thought the clock was good enough? Like what was your you know how did you decide okay now we're going to do the three-year measurement? >> I don't know. I I think I think you went you you ran the experiment as long as you could stand it.

U I mean >> Oh no I don't mean why did you stop at 3 years? I mean you why did you start like let's say what year I don't know what year you started the thing >> 1955 >> okay so in 1955 how did you know that the clock was good enough to start like the long experiment without making any >> I didn't know how to make it any better >> how did you know that it was how were you like measuring its goodness like what was the >> I built two of them and compared them >> okay okay interesting >> okay I built two of them and compared them >> and so you would tweak one and see if it improved the the repeatability or something like that >> and I I I I sat down and I What are the problems?

I mean, you understand it wasn't me, okay? It it was it was it was R, you know, uh it was Perry, okay, and so on. And and and uh uh you sit down and you think about you make it as clever as you can and then you build two of them. >> And so you'd make a tweak in one and that would be version 1.10 or whatever 1.1. >> Well, and you play you make it as best as you can and then you say, "This is best as I know how to make it.

It's time to start the experiment." And the figure of merit was like the stability of the frequency that you were uh so like dayto-day what's the stability. >> Yeah, it was like a part of nine. >> Okay. >> It it was by by current standards it was very crude. >> Mhm. >> It was it was like a part 109 part of 10. It was it was several orders of magnitude poorer than than the ones today. There's been a lot of engineering developments to make things better.

So at that time, how long did it or like the whole you or the whole team to to build the the cium clock? >> Like how how long does it take like to build? >> How many people? Yeah. How many people were were building it and how long did it take? >> Uh there were two of them and they were in the UK. Uh >> that was the first one, right? >> Yeah. Was >> I guess maybe here what you you were involved in. >> Oh no, no, no. It was way before me. >> Yeah. >> Oh, it was way before me.

No, the one you built. Did you build the clock? >> The by the time I built the clocks, they were the engineering was well understood. I'm not sure I understand. >> No, the question is just what when you were building the clock that ended up, you know, creating UTC. How long? >> Oh, no. The clocks the clocks that are create UTC are commercial devices. I >> Okay. So, so, so maybe tell us exactly what you were doing in the 50s.

Like what was what did it look like? >> No, not in the 50s. was in seven years. >> It was the 60s. It was 60 years. >> The answer is understanding the statistics. >> Okay, >> that is because because the clocks are imperfect and and you have to model them if you have if you have an ensemble of clocks. They don't they don't they all disagree what the time is. And the question is what's the best estimate you have of what the time is given that you have a bunch of clocks that don't tell you the same answer.

So um like especially with recent advances in uh like machine learning and AI kind of data science approaches do do you see the way to uh kind of improve the like collecting the data on the on the clock to improve the performance >> AI AI doesn't >> I would say maybe data science so you collect a lot of data but then you would try to feed to optimize the performance of the clock >> yeah I think that that I think that's right that's that's the right idea Yeah.

Okay. But >> since each of the clock it has like its own personality, its own >> you have >> Oh, here we go. >> Okay. The answer is you have frequency stability of the clocks and type stability of clocks >> and and and you have different clocks and you try and model when you have an ensemble of clocks, you try and model the behavior. I mean that that's that's just called the elevator. It's just a way of doing it. But but but but conceptually you have you you have you have an ensemble of clocks that don't agree with each other and you have to try and estimate what's the best information that I have from the ensemble given that the ensemble does not agree that the diff the individual members of the ensemble do not agree with each other.

So I have it's so it's it's a it's a it's basically a statistics problem >> but but it's a it's a statistics problem of a style that was not understood beforehand. That is it was it was uh it was not that was one of the primary developments of the era was understanding the statistics of atomic clocks and how you how you take an ensemble average and get the best you can out of out of the ensemble clocks. What's the what were you doing besides literally if you have 10 clocks just averaging them like what's the >> no no averaging averaging is not is not good enough and the reason averaging is not good enough is that the clocks don't have noise they have systematic variations >> that is there is clock number one in which the manufacturing tolerances are that it's magnetic it stray magnetic field is a little larger >> and so averaging that clock does not help you get the wrong time after time after time >> timing inaccuracy. >> Okay, it's an accuracy problem. >> So, so you have to you have to estimate the the systematic errors as well as the random errors. >> Okay, and the systematic errors in the end eat your lunch. >> But how can you how do you know the systematic errors?

Let's say you have 10 clocks. You don't know which one is right. >> The answer is I I I I learn by by by sending the data to the BIPM and comparing it with other laboratories. And we have a community understanding of the systematic errors. So the BIPM provides a community understanding. >> And so basically in order to set the time, let's say you have 10 clocks and you have a formula like each clock contributes its own little bit to the formula and you say okay given that that this is the time on all these 10 clocks and here are my fudge factors for each clock.

This is what we think the time is. >> Right. >> Okay. Okay. And now now I take that time and I send it to the BIPM and the BIPM says you're two nanoconds fast with respect to everybody else >> or you're three nan slow with respect to everybody else or in fact what they send me is a table that that every five days they tell me where I am with respect to everybody else. Okay. And they say I'm two nc on last week you were two nconds fast this week you're three nconds fast.

So there's partially an assumption of this kind of you're essentially manufacturing reality by saying like okay we're going to have multiple places where everybody tries to come up with you know reality and then we're going to compare it with each other and try to guess at like you know we don't know the hard truth but given that everybody's sources of error should be different. >> What do you mean by hard truth? like because there's no clock, you know, like there's no like actual it's all it's all clocks with errors, right?

You're basically saying like we think the different labs have different sources of errors. So if like let's say three out of four labs are slow versus the fast one, the fast one is probably has an error. >> Okay, that should be >> but what you what you have come to the idea is that time is a human construction. That seems to be the case based on this discussion. >> Time I mean Yeah. time is a human construction. Yeah.

When I say it's 4:00 or 20 after 4 or whatever the time is, it's it's a human construction. And and we all agree that that when you say you're going to be here at 2:00, we all agree what 2:00 means. >> We synchronize. >> We synchronize the right place at the right time. >> And it it does it doesn't matter what we call it. >> All right. >> Okay. Well, we all agree that that you're going to show up here at 2 o'clock and we all understand what 2 o'clock means. >> Okay.

But but but there is a a platonic time. We just can't measure it or no. >> I don't know what that means even >> like a theoretical time. Some like kind of like math, you know, like 2 plus 2 is four even if the nothing else exist. >> But but but but but time does not have that quality. Time is a human construction. >> I think like it's all in terms of synchronization like object A gets into point B. Well, but but but but the only reason the most important aspect of time is that it's a way of so that everything doesn't happen at once. >> But do do you think math is a human construction? >> Math like just just you know summation or something like that.

Does that exist? >> But math and time are not the same thing. >> No, no, I know. I agree. I'm just saying do you just to set a baseline like do you think math exists outside of humans inventing I think I think algebra exists outside of human construction but time does not. >> Okay. Okay. >> Time is a human construction. It it is simply it is simply that that that it is a way of organizing society such that that uh that that we look at the clock and we all agree to do the same thing at the same time and but we we instead of calling at 2 o'clock we could have called it 3:00 or 8:00 or 12:00 or whatever it is.

So, so let's say you have all these physical things happening, you know, atoms moving around, chemical reactions, whatever the entire universe, everything happens with respect to time, whatever that is. >> Well, you got you got to you have to be careful to distinguish between time and time interval. >> Okay. >> Time interval is different. >> Time interval you think is real. >> Time interval is is time interval is is is basically a number of cycles of cesium and is real. >> Yeah. time interval is real. >> Do you think that if there was like a drift in that in the universe we like would we see it in the work that >> N there are people who are looking for that >> the answer is okay one way of looking for it is let me look at the light from a very distant star. >> Mhm. >> Okay.

Now the light from a very distant star left the star a long long long long time time ago. All right. Do I see something funny about the light from a distant star? The answer is I don't know. Not yet. But if I see something funny from a light from a distant star, I've addressed your question. >> But in clock systems, there is there is there no chance because they're they're like kind of like circular reasoning. >> Well, I I think the the answer is is a question of magnitude. >> Uh because they're operating for such a short time. >> Well, I mean the the the the stiffest test that you've talked about are to look at the light from a distant star. because it's been just existed for a very long time. >> Because it because it left the star a very very long time ago, >> right?

I >> mean that those are the stiffest tests. >> It's like a latency delay. >> Was that you mean the the the for the light like trouble from the star? >> Yeah. Yeah. >> Yeah. Well, it left a delay. It left the star billions of years ago. >> Is there nothing else like that? Like if you just pick up a an object, right? Does it contain nothing that like records time? >> No. You you got to always distinguish between time and time interval, >> right? >> Okay.

So the answer is >> like the passage of time. Let's say uranium decays over time, right? Stuff like that. >> All all atoms have implicit time interval, implicit frequencies, >> right? Right. Right. >> But that's that's not time. That's time interval. >> It's like cycles like cycles. So let's say if the frequency of the cesium shift change of the cesium shift was subtly changing over like you know as the universe got gets older would you guys detect that or would that be kind of cancel out and all >> I think the the answer maybe and maybe not because >> in order for me to detect that I need to have something to back up against that does not change >> right >> so now you have to identify the c it's changing in the cesium frequency but somehow everything else isn't changing or something else something else not changing Right.

Right. Right. >> And and I don't know. Okay. I mean for example the question is are there atomic properties that are changing in a way that changes because the cium frequency is derived from atomic properties or maybe one of the atomic properties is changing. Maybe the mass of the atom is changing or maybe the charge whatever. And there are people looking at those kind of questions. >> Okay. But those are atomic physics questions.

Those those are not really clock questions. >> But I guess you could use a clock to probe some of that because it's the most precise. >> That's right. And and there are a lot of experiments which look for those kind of experiment and the answer is n there's not there isn't there isn't there isn't a systematic there are tantalizing evidence that may be let me say it that way >> okay what's an example of that >> uh there's a thing called the fine structure constant which may be changing >> the fine structure constant is is a combination of of atomic properties uh which turns out to be a real number What of 137 it just turns out to be an it turns out to be a well-known fraction and maybe it's changing maybe not the data are very poor they are very noisy >> it's interesting that that uh uh alpha uh it was like defined in a very kind of dirty experiment like that for >> the variation of the price of constant is an active experiment I mean >> it's the the quantum hall effect >> I mean the idea is to find is to find something which varies with the fine structure constant differently from the way something else varies with the fine structure constant and then see if those two those two measures of the fine structure constant diverge from each other and and I'm not an expert on fine structure constants but I don't think so I don't think the data are definitive the data are suggestive >> speaking of time like time time intervals so I see now it's more like like in let's say in quantum mechanics so like in atoms it's or it's more cycles like how many transitions or how often do they happen?

But then if we look at the uh like general theory of relativity, okay, >> those like the the space-time continuums, >> how do we bring this into >> Well, I think I think general relativity has an impact that that what you observe a clock depends on where you are, >> right? and where the what your gravitational potential is and are you moving with respect to the clock or is the clock moving with respect to you or is the clock at a different gravitational potential and and and uh uh and so general relativity has has a has a lot to say about that but general relativity doesn't affect what happens if I have a clock right here in my hand you know that that's that's my local time that's the proper time of the clock and the proper time of the clock I I don't I mean, I'm moving with the clock.

Every test that I have doesn't detect that effect. >> Could clock be one of this the stitching points between >> Say it again. >> Could like like time definition or clocking could be like the defining like the stitching point between like general relativity and the quantum mechanics because it seems like what you define that time it's more on the transition of the atom. >> I don't know. I I don't I don't know. I I don't know the answer to that.

I don't think so. But I don't know. Okay. The question is whether whether relativity is applied to quantum mechanics and I don't think so. Uh I don't know. I'm not that's something I really don't know about. >> Also for let's say young scientists coming to the field of um like let's say clocks or precision measurements. So uh there have been tremendous progress let's say building the atomic clocks. Now we can use like this let's say optical atomic clocks like go to precision like 10 tous 19.

What is left like what uh what are the next biggest challenges you see or like the problems unsolved? Well, I think I think there's a you got to remember that that that having a clock in your laboratory is very exciting, but using the clock is presents all sorts of challenges. Uh, and that has to do with with characterizing channel delays and and 5G or 6G or whatever or or or synchronizing widely spaced clocks or widely spaced generators and and those are not trivial issues.

It's it's it's difficult it's difficult to exploit the full capability of atomic clocks uh with in remote dis with remote synchronization. It's possible, but it's very difficult. That is it's it's it's very hard if if you and I are separated by by a continent or some fraction of a continent. It's very hard for me to tell you the time with the accuracy that I know it locally. that is I know the time locally but it's very difficult for me to tell you the time uh in New York City with the same accuracy that I know it because the transmission delay and the knowledge of the transmission delay and the characteristics of the channel and all that stuff uh are very important and in fact they they uh in mo in most situations they dominate the problem so that that having a clock here u I mean unless I have another clock in the same room. >> Uh I can't necessarily exploit >> the advance.

For example, it's very difficult if June's clock would have worked perfectly. It would be very difficult to exploit it because it would be very difficult to transmit the time with the accuracy that June knows it. >> I see. So that's the kind of the communication channel problem. >> It's a channel problem. It's a receiving problem. It's a signal and noise problem. There are all sorts of problems, but the channel delay is one of them.

Yeah, channel delay is an important problem. >> Are there any ideas on the horizon how to improve the the channel problems? >> Uh well, there are there are dedicated fiber circuits. Uh there are uh uh but but but the main problem with the channel problem is that it's is that good channels are expensive and they tend to be pointtooint. that is uh uh it would be it would be possible for me for you and I to transmit time to each other on a dedicated channel.

It would be expensive but we might be able to do it technically. >> Mhm. >> But it's a pointto-point transmission. Uh namely uh if I now wanted to transmit time to my to your buddy, I now got to repeat the whole thing all over again. Okay, that is uh it is not it's not so useful. So we've made point measurements but we haven't made a global measurement. But the transmission problem is a very serious problem. Uh and and and uh uh it's it's even with a even without optical clocks, it's very difficult to transmit the time to a remote user as well as it's known locally.

It's possible but it's very difficult. >> I see. So that clocks maybe with high accuracy they could be used for maybe as some sort of like sensors >> maybe I >> but not but maybe not as like a indistributive network. >> Well I I think I think that the the development of clocks has outstripped the development of transmission medium. That's I think that's that's a better way to say. >> So it's kind of behind lagging behind. >> Yeah. is lagging behind and and and so so so uh an important an important development is to look not at the improvement of clocks but the improvement of transmissions that that is a very important aspect >> is it more of like an engineering type of problem like materials >> fiber yeah well okay but it it's also >> uh it's also a matter of principle uh in the sense that yeah engineering is important But but let's talk about the physics.

I mean, how are we going to measure the delay? >> Yeah. >> And and and what what mechanism am I going to use for measuring the delay and how accurately can I do it? And and and uh uh and so now after I decide on that question, I have to decide on the engineering questions. Okay. How do I actually build it? Uh and and and that's that's another issue. Okay. But but uh but I've given you the method. I mean the two-way method, the common method, those are the ways.

And now maybe there's a clever way that we haven't thought of yet, but those are the ways. >> Are there any driving use cases like driving this technology or we're like more or less happy with >> Oh, no, no, no, no. There's always there's always a driving use case. Okay. There's always Okay. There's always there always people who want more. >> Yeah. >> Okay. There's 5G, there's 6G, there's there's always people who want more. >> How much do we need for 6G?

What's the >> requirements? I think 6G I don't I don't know. I'm not a telecom engineer, but I think 6 feet could probably accommodated uh with the existing clocks, but I'm not sure about the transmission medium. I I'm not sure about that. I I'm not but I don't I'm not sure that's that's an authoritative answer. I don't know. >> I see. Uh for GPS, I guess GPS is >> GPS GPS uses commercial clocks, uses rubidiums and cesiums. >> I see.

Now uh the next generation of of the GPS like satellites may use uh hydrogen mazers or maybe optical clocks if they would if they were reliable. I mean the trouble with optical clocks is the moment they're not reliable enough. You can't put them up in a satellite because they don't have a they don't have a long lifetime. >> What's the advantage of hydrogen masers compared to >> they're very they have very good short-term stability. >> The their short-term stability is the best of all of the commercial devices.

I see. And the global navigation system I guess is more like high precision. Is it I assume like more for military purposes? I don't think for a civil >> Well, I think there there u I mean there certainly certainly there are military applications but the the GPS system is used for synchronizing telecom cell sites and so on. It has uh it has commercial applications. I mean, it's it's the backbone of the cell system. >> What about the the Starlink?

Do they need the better? >> I don't Okay. I I don't know. I don't know. I don't know if Star I don't know if if I don't know if are the satellites in Star League active or passive? >> Active. >> They're active. There we go. >> So So they're active. But but the question is what what kind of clocks do they have on the board? I mean, I don't know what they have. I don't know. I mean I I just don't know >> because it seems like they the company probably has most of the satellites >> isn't doing it's not really GPS it's more like uh it's data transmission right transmission.

Exactly. >> I think it's mostly data transmission. Yeah. I I think the the most of the satellites use rubidium clocks because the reliability is best. the the uh the GPS system I think uses mostly rubidium clocks. >> Why rubidium? >> Because the reliability is better.