Oral-History:Judah Levine
About Judah Levine
Judah Levine was born in New York City in 1940. He attended Yeshiva College and received a degree of Bachelor of Arts in 1960 with a major in physics and a minor in mathematics. He then went to graduate school at New York University, and was awarded the degree of Master of Science in 1962 and Doctor of Philosophy in 1966. He visited the Clarendon Laboratory of Oxford University as a post-doc from 1966-67 and then was a postdoc at JILA, an institute operated jointly by the University of Colorado and NBS/NIST from 1967-1969. He joined the National Bureau of Standards in 1969, where he worked on various applications of frequency-stabilized lasers including laser radar systems and a 30 m interferometer that he used for various geophysical studies. He moved to the Time and Frequency division of NBS in 1972 and has worked in many areas of time and frequency since that time. His work in includes the design and realization of the NBS/NIST time scale that is used to compute UTC(NIST) and a number of applications to distribute time and frequency information to users by various digital methods. He is currently a Fellow of NIST and an Adjoint Professor in the University of Colorado, Department of Physics. He has received the Colorado Governor’s award for Research in Innovation Technology, the I. I. Rabi award of the IEEE UFFC Frequency Control Symposium, and two Department of Commerce Gold Medals. As of 2021, he is a staff member of the NIST Time and Frequency Division, and he works on problems of time scale algorithms, the definition of UTC, and digital methods for transmitting time information.
About the Interview
Judah Levine: An interview conducted by John Kitching for the IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society and for the IEEE History Center on Friday, 11 October, 2019 in John Kitching’s Office at NIST, 325 S. Broadway, Boulder, CO 80305.
Interview #851 For the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.
Copyright Statement
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.
It is recommended that this oral history be cited as follows:
Judah Levine, an oral history conducted in 2019 by John Kitching, IEEE History Center, Piscataway, NJ, USA.
Interview
Interviewer: John Kitching
Interviewee: Judah Levine
Date: October 11th, 2019
Location: Boulder, CO
Kitching:
My name is John Kitching and I'm here with Judah Levine at the NIST Boulder Laboratories on October 11th, 2019. Judah has worked in the field of time and frequency for several decades and among other things pioneered the NIST Internet Time Service. I'm delighted to be able to do an oral history of his career and life for the IEEE-UFFC. I'd like to start with your early life. Can you tell me where and when you were born?
Levine:
I was born in 1940 in the Bronx, which is a part of New York City. My father was an elementary school teacher. He taught the sixth grade in the New York City public school system. He always had at least one additional job during the school year, and usually worked in a summer camp or taught in the summer school program as well. My mother stayed home pretty much most of the time. We lived in what was called a “project”, which is sort of the lower end of rent-controlled “affordable” housing. My mother's family came from Jerusalem and she came to the United States in about 1910 or so. She was 3 or 4 years old at the time. My father was born in the United States, but just barely, because his older brother and his older sister were born in Europe. They came from Eastern Europe, from Poland. That's sort of typical of the Jews who came to New York city in the 1900 - 1910 era.
Levine:
- Audio File
- MP3 Audio
(Levine clip 1.mp3)
My mother's family came from a long line of rabbis and scholars, so the family decided or hoped or whatever, that I would become a rabbi as well. And so from the earliest time, I was enrolled in a Jewish day school, which is an all-day school in which roughly 50 percent of the time is on Jewish studies, the Bible and Talmud and language and so on, and the other half of the time is on the usual English curriculum. A school day was typically from 9 am to 5:30 pm on Sunday through Thursday, and only from 9 am to 1 pm on Friday, so that there would be enough time to prepare for the Sabbath, which began at sunset. I continued with this schedule from the first grade until my last two years in college, when I no longer had a Talmud class on Sunday morning.
Kitching:
You went to this school through which grade?
Levine:
I went to that school for my entire education. I went to a series of very-similar schools and I graduated from Yeshiva High School in 1956 and from Yeshiva College in 1960. Yeshiva College, of course, was the same program. There's a half a day of, by then it was advanced Talmud study, and a half a day of typical college stuff. It's a pretty full schedule. You kind of spend from nine to six every day in class. Originally, it was six days a week, five and a half days a week. Then it was five days a week. I was in the pre-rabbinical program, but I also majored in physics and as I got closer to graduation, I realized that I was never going to go into the rabbi business. In spite of what my family wanted. I graduated in 1960 with a major in physics and a minor in mathematics from Yeshiva College and with several years of advanced Talmud study. If I had stayed around another year or two, I would have become an ordained rabbi, but I never did that.
Kitching:
Let me ask you; has religion and your Jewish heritage played a role in your life?
Levine:
Oh yes! I was much more observant as a young person. Both my families, all my families, were very observant. I became less and less observant as I got older. I'm a member of the Reform Congregation here [in Boulder]. From time to time, I've taught classes in the Talmud and Bible in Hebrew and so on. I am currently (November 2020) a member of a zoom-based discussion group on Medieval Jewish Philosophy. Most of the members of the congregation really have no knowledge of the Jewish heritage or the Jewish background. I am something of an anomaly in that I come from a much more traditional background than almost everybody else in the congregation. I'm not particularly observant anymore. I observe a few things but not very much. My wife is about the same. We both come from very traditional backgrounds and we both observe a lot less.
Kitching:
I want to come back to this maybe later in the interview, but I'm going to go back to your early life. What did you do for fun as a child growing up?
Levine:
You understand that if you're going to school from 9 am to 6 pm, six days a week, there isn't a lot of time for fun. The only day when you don't go to school is on the Sabbath. On the Sabbath, traditional Jews basically can't do anything. That is, you can't travel, you can't do anything that is technically called work, but which we wouldn't think of as work. We would think of a recreational thing as “not work.” The traditional view is that almost anything that you do is “work.”. Carrying money is work. Traveling is work. You sort of can't do much. I went to services on Saturday morning, and Saturday afternoon was typically a time for meeting friends in Bronx Park. You can't travel to Bronx Park, but I could walk to Bronx Park.
We all met at the seal pool in the middle of Bronx park. I thought the Bronx Zoo was really a very fine place, in spite of the fact that the Bronx itself is, sort of, not exactly as wonderful now as it was when I was a child. At the time, the neighborhood I lived in was about evenly divided between lower middle class first and second-generation Jews, Irish, and Italians. Time has not been kind to that area of the Bronx. I guess I'm not sure that is characterized as fun, but that's about all there was.
Kitching:
What were your favorite topics in school as a child? Were you already interested in technical subjects at that time? Levine Oh yes. I was interested in two things: I was interested in electricity, just in general. Nobody in my family knew anything about what it was. I found all these old books. This was in the 1950s, let's say, when I was ten. The books had been written in the 1930s and I had no guidance. I had nobody to help me to understand it. I didn't know what I was doing really, and I just played around. I would do things that scared my parents silly because they didn't understand what was going on.
Kitching:
It wasn't just books, it was also a playing around with electricity.
Levine:
Yes. I would wire things up, and one day I managed to blow the fuse in the whole building and all of the lights to the whole building went out. This is an apartment house and I'm not quite sure how I managed to do that. My parents were mortified that I had done that, but nobody quite knew what had happened. The only thing that people saw was that all of a sudden all the lights in the building went out.
Kitching:
They probably suspected you.
Levine:
I don't know. I don't think they suspected me. I didn't think they knew what had happened. You have to remember that this is a New York City lower-income project and things happened. If the power went out (it didn't go out normally, but it went out), the idea that this was just business as usual was not totally impossible. The other thing I was interested in was I had an elementary chemistry set and I played around with the chemistry set, but again, the same thing, I didn’t really know what I was doing because I had no guidance, no help, no books, no nothing. I just tried things and I managed to not set the house on fire, but it was close.
Kitching:
How is it that you developed your interest in math and physics without, it seems, any mentorship at all?
Levine:
I guess the focus on physics really started when I got to college because in high school there really was no opportunity for any sort of specialization. Many of the opportunities that are available to my kids, for example in the Boulder Public Schools were not available. They just didn't have that kind of stuff. In college, I guess the choices were physics or chemistry. I could have stayed in what you might call the rabbi track, but I wasn't interested in that. I guess I came to physics as much by default as anything else Electricity was a really interesting point and that turned out to be very useful in the undergraduate laboratories. I don't know, I don't think there was a lot of choice for me. In my last two years of college, I got a job as an assistant in the Physics department laboratories, setting up experiments and repairing equipment. I probably would not have been able to get that job without my previous experience in playing around with stuff. This was very important to me because I would have had great difficulty staying in college without this job.
Kitching:
Was it obvious to you as you were finishing high school that you were going to go to college?
Levine:
Oh yes. Not only was I going to go to college, but also I was going to go to rabbinical school. There was no discussion about that. You're going to go to Yeshiva College and you're going to be ordained. That was what the family was ready for. Yeshiva College was regarded as a compromise because it had a serious English curriculum and they would just as soon have minimized the English curriculum because being a rabbi is being a rabbi and the religious aspects of the curriculum are much more important. On the other hand, I don't think they were like some of the more extreme versions of this idea in which you don't study English stuff at all. They weren't prepared for that because most of my mother's brothers were ordained rabbis, but not practicing rabbis. You know they didn't have a congregation. They would do whatever they did. One of my uncles was a stockbroker. One of them was an architect. They were all ordained rabbis, but they weren't practicing rabbis. I think that was the model that they had for me.
Kitching:
Did anyone recognize your clear abilities in physics and math early on?
Levine:
I guess the earliest thing about that was that when I was in college, I got what we would call here a teaching assistantship to run the undergraduate laboratory. That was in physics. Over at CU there's a professional guy who does that full time, but at Yeshiva College, I was the guy who set up the laboratory experiments for the undergraduate classes. I maintained the equipment. Part of it was because I was interested in physics and part of it because I couldn't afford to go to college. I couldn't afford the tuition without some kind of teaching assistantship. When I went to graduate school, it was the same thing. I had a teaching assistantship in which I set up the lecture demonstrations, so I maintained the lecture demonstrations for all of the undergraduate classes. The advantage of that teaching assistantship was that it was a twelve-month TA and that was very important because I probably couldn't have done it without that. I also had a night job as a switchboard operator at a hospital on 107th Street in Harlem. The place was pretty quiet after the day staff left and visiting hours were over, so it was a good way to get paid to do homework.
Kitching:
What did you do in the summers when you were in college?
Levine:
You've got to remember that I'm going to school from 9 am to 6 pm, five days a week. By the time you get home, boy, you're really done. Saturday is the day of rest, and Sunday is the day of homework.
In the summer, my parents went to the Catskill Mountains. We rented a bungalow. I don't know if that concept still exists, but you rent the bungalow from Memorial Day to Labor Day. You're out in the fresh air and you go to a thing that amounts to a day camp and this is the usual day camp kind of activities. I went swimming almost every day, mostly in Lake Huntington, which was in a nearby town. The town also had a bowling alley with a person who set up the pins. In later years, there was a small movie theater in back of the drug store.
Kitching:
You did your undergraduate degree at Yeshiva College, and as you were finishing that, you decided to go to graduate school?
Levine:
Oh, yes.
Kitching:
Where did you go to graduate school?
Levine:
I went to New York University and that was also a compromise because I really wanted to go to Columbia. I had applied to Columbia, but they did not give me an assistantship, so I couldn't afford to go. NYU gave me a teaching assistantship where I set up the lecture demonstrations and that's what decided it.
I went there because that was the only place I could afford to go. The fact that it was a twelve-month teaching assistantship was important, though I didn't know that that's what would happen at the time. I didn't know that it was going to be a twelve-month assistantship. I figured that in the summertime, I'd have to do something else, but in fact, it was a twelve-month job, full time.
Kitching:
What was your thesis topic?
Levine:
New York University at the time had two campuses. There was the University Heights Campus and a Washington Square Campus. I was at the University Heights Campus. The best research program at the time at the University Heights Campus was in atomic physics, in atomic beams.
My thesis advisor was Ben Bederson. I did a thesis research topic on the polarizability of meta-stable mercury, which was an atomic beam concept. [Ben Bederson] had a pretty large group. He had a group of about fifteen students. The way the group was organized was that you were the junior assistant to somebody else who was getting a degree. Then when that person got the degree, you inherited the apparatus, and then you did something else with the same apparatus. That's what happened to me.
My “fathers” were Ed Pollock, who went to University of Connecticut at Storrs and Ed Robinson, who stayed on the faculty of NYU. My “son” was Bob Celotta, who worked at NIST/Gaithersburg for many years. I graduated; I got a Ph.D. in 1966 in atomic physics. I went to the classes, the research project was up at University Heights, but the classes were mostly down in Washington Square. I would travel back and forth between the two campuses; the classes were mostly at night.
Kitching:
At that time, I guess the group at Columbia, formed probably by I. I. Rabi and others…
Levine:
Yes, it was a little past Rabi’s time. I knew Rabi, but he had sort of retired by then. All the famous guys, Rabi and [Polycarp] Kusch and so on, it was a little bit after that time. The Radiation Laboratory was still the place to be in Columbia, but the decision was purely economic. I simply couldn't afford to go. A number of my friends went to Columbia, and in fact, they thought it was terrible in the end. Maybe that's not something for publication, but they thought that the graduate students were treated very poorly. That was their opinion. Graduate students are not treated wonderfully anywhere. As I've gotten older, I've felt more and more respect for Ben Bederson, and at the time, you sort of think, eh. When you think back on it, he was really a good guy. He really did the right thing.
Kitching:
What do you mean by that specifically?
Levine:
He was reasonable. He was fair. He did what many thesis advisors do, which was to suggest the thesis topic, which was within 10 percent of being totally impossible. We said, I understand that's a difficult one, so I'm gonna just do a warm up project and you should do this in a weekend or two and that's the polarizability of metastable mercury and yes, that's just an easy one.
That took four years because you see my original thesis topic was going to be the polarizability of atomic hydrogen in the ground state. The reason that's interesting is that that's a calculable thing. Hydrogen is so simple you can calculate it. However, from the experimental perspective, that's a profoundly difficult experiment, at least in the atomic beams concept, because it’s hard to make atomic hydrogen, it’s hard to detect atomic hydrogen. The polarizability is very small, which meant that very large electric fields would have been required. It has all sorts of technical issues. It's a really difficult issue. The only advantage at least at that time was that you could calculate it, exactly. You knew exactly what the DC polarizability was going to be, whereas with metastable mercury; it's a long, involved story, but you could never calculate exactly what the polarizability was. There are two low-lying metastable states in 6s6p 3P mercury with different values of the angular quantum number, J, and calculating the J- and mJ dependence of the polarizability had significant uncertainties, given the theory at that time. In fact, the results that I got did not quite agree with theory (especially with the magnitude of the tensor contribution) and maybe it was a problem at their end; it was sort of hard to tell. There is also a high-lying metastable state that is easier to detect but harder to produce, and I was never able to get a good measurement of the polarizability of that state.
Kitching:
Why did this topic interest you? Was it mainly because your advisor suggested it that you went down that road or did you have a particular enthusiasm for this type of experiment?
Levine:
Once you are at university, at NYU University Heights, that's it, the only really first-class experimental research project there at the time was Ben Bederson in atomic physics. As I said, he had a number of students and I guess you might say he put me in the pipeline. You're the junior student. I didn't understand about the polarizability of metastable mercury. Measuring the polarizability of metastable mercury was a natural extension of the program in Ben’s lab to measure polarizabilities. The previous students had measured the polarizabilities of all of the alkali atoms and the metastable rare gasses, but I didn’t have that perspective. I’m a young graduate student. It's only when I found out how difficult that was going to be that I said, “Holy Moly.” I managed to measure the polarizability with the first computer-controlled experiment in Ben’s group. The “computer” was a monstrosity that had numerous vacuum tubes and four hundred words of core memory – real magnetic cores. Nobody in Ben’s group had seen anything like it. I was able to measure the polarizabilities of metastable helium and metastable argon in a few hours – something that had taken months by the previous manual methods.
I don't think I had a lot of choice of thesis topics. That was what it was going to be. I think, when I came out, I knew a lot of stuff. Atomic beams is a technically challenging area. Maybe it's not so much anymore, but for example, it's the basis for all of the primary frequency standards. They're all atomic beam experiments. By the time I came to NIST, I knew pretty much everything there was to know about atomic beam stuff, much more than most of the people at NIST at the time who were dealing with the primary frequency standards.
Kitching:
How often did you meet with your advisor? He had a big group, it sounded like. Did you mainly learn from other students?
Levine:
Ben had this idea that what you do is you throw somebody in the deep end of the pool and you walk away, so he did that. He threw you in the deep by the pool and he walked away. He would come around periodically, but we didn't have anything like group meetings. We never had a group meeting. Periodically you would be summoned to “the presence” and he would want to know what's going on, blah, blah, blah. Actually, Ben spent a lot of time as a visiting fellow with JILA, so he was gone a lot. When he was gone, you did what you did. On the other hand, he had like fifteen graduate students and we were all in two connected laboratories; there were two large rooms that were connected together. You learn from other students and the other students teach you and I think that's a good system in the end. You're the junior student on a lot of things, so you'll learn a lot of stuff just from the fact that you're working for another student who's on the final approach, so to speak.
I have not dealt with my students that way. I've been much more hands-on with my students, which means I only have one or two at a time because I really devote a lot of time to them, much more than Ben did to me. Ben sort of just said “here it is, you know, go do something”.
Kitching:
Why have you changed that approach; was that intentional? Often, when we do a Ph.D., we learn from our advisors how to advise, but it sounds like you've taken a very different tack.
Levine:
When you have kids, either you do it exactly the way your parents did it or exactly not the way your parents did it. I did exactly not the way my parents did. I treated students in a much more interactive, hands-on kind of way that you could never do if you had fifteen of them. You just couldn't bother with it. I don't know if that's good or that's bad. It is what it is. I've never had more than three students at a time and usually it's two.
Kitching:
You graduated in 1966 with your Ph.D. and your advisor spent a lot of time out here at JILA. What happened then?
Levine:
What happened then was that NYU had a colloquium series where they invited great people to come and give colloquia. One of the guys they invited was Dirac. He was an old man at the time. I remember him clearly.
Kitching:
This was Paul Dirac?
Levine:
Paul Dirac. Yes. I had the idea that he was French, but in fact, he spoke with a pretty heavy English accent. When he came and gave a colloquium, I decided that I was going to become a theorist because he explained things in a way that even I, as a relatively young graduate student, understood everything, or at least I thought I understood everything. Then the next week, or maybe a week after, Bob Dicke from Princeton gave the colloquium.
This is New York City, so everybody's around. [Dicke] gave the colloquium and at the time, there was an alternative to general relativity called the Brans-Dicke theory. One of the tests for relativity is the advance of the perihelion of Mercury. Dicke proposed that the advance of the perihelion of Mercury was really due to a quadrupole moment in the sun. He was doing an experiment to measure the quadruple moment of the sun. The oblateness of the sun. He gave a talk on that that was just infinitely wonderful. I said, that's where I want to go. I want to go to Princeton to work with Dicke as a postdoc.
In those days, you didn't just apply for a postdoc. You had your advisor discuss it with the person, so I asked Ben to send a letter to Dicke to ask him if he would take me as a postdoc. What I got back was a one-line refusal, which just said, “I don't accept students who’ve been trained in atomic physics.” I said, “Oh, well, okay.”
Ten Ben came back and said “Why don't you go to Oxford and work with Pat Sandars?” Pat Sandars at the time was looking for the permanent dipole moment of the electron, which is still rattling around in JILA. Eric [Cornell] is still doing it. [Ben said,] “I'll get you a postdoc at Oxford and you go to Oxford for a year and you work on the electric dipole moment.” That was not as good as the oblateness of the sun, but it was close. Pat’s proposal was to look for a permanent electric dipole moment in metastable xenon in an atomic-beam configuration. I knew all about that [because] I had been working with metastable rare-gas beams for years, so I built a beam apparatus starting from nothing. There was no way of finishing the experiment in a year, so I had to leave just when the apparatus was starting to work.
Kitching:
It was looking for the oblateness of the electron, I guess!
Levine:
Effectively, yes, or the asymmetry of the electron. At the time, Ed Robinson, who was the student who was ahead of me on my apparatus had come to JILA as a postdoc.
On Christmas of 1964 when I was still a graduate student, he invited me to come out and spend Christmas with him in Boulder, so I came out and spent Christmas in Boulder. At that time, JILA was not in the building it’s in now. It was still back in the armory. I met the various people in JILA and all that, but I was just a student. I met Dick Zare, who was I think a visiting fellow or maybe a junior faculty member. Then in 1966 when I graduated from NYU, I went to England for a year and halfway through my year I said, “Oops, I gotta start looking for a job,” because the postdoc was a year, and then “goodbye,” that was it. Looking for a job from England is hard, because there's a time zone problem.
Then one day I got a telegram from Lou Branscomb who was at the time the head of JILA and he said, “meet me in London. I'm going to offer you a postdoc.” This was completely out of the blue. I said, “all right, well, you know,” and I took the train to London. He offered me a postdoc. He said “Judah, I’m going to offer you an ARPA postdoc, which means that you come to JILA and you do what you like. This is just, you know, you sit in your office, you think great thoughts, and your nominal supervisor will be Jan Hall.”
Kitching:
You said an ARPA postdoc?
Levine:
That was what it was called at the time. JILA had a large block grant from ARPA.
Kitching:
From the Advanced Research Projects Agency, that eventually became DARPA, the Defense Advanced Research Projects Agency.
Levine:
Yes. ARPA had this program for postdocs. I came to JILA in the fall of 1967, I was an ARPA postdoc, and I started working with Jan Hall. However, it turned out that it wasn't quite what Branscomb had in mind. At the time, ARPA was interested in the physics of the upper atmosphere, for various reasons. One of the interesting issues was the electron affinity of atmospheric gases, and that had to do with electromagnetic pulses and it doesn't really matter, but there were all these issues about what happens if you set off a bomb high in the atmosphere? You get this cloud of electrons and so on. One of the issues was what's the electron affinity of various atoms.
Jan was working on the electron affinity of O2-. Which was what ARPA was interested in at the time. In fact, almost everybody in JILA was working on that kind of stuff. I came and started working on the electron affinity of O2-. Like everything else, that turned out to be an incredibly difficult experiment technically, even though it was supposed to be a weekend experiment. The experiment was to produce a beam of O2-, detach the electron with photons from an argon-ion laser at 488 nm, and then measure the energy of the liberated electrons.
It was one of these things you were going to do on the weekend and move on to something else, but after two years, Jan and I still had not quite done it. We knew what the answer was, but we couldn't quite get the right answer. The problem turned out to be a very strong dependence of the photo-detachment cross-section on the polarization of the laser beam. In those days, a JILA postdoc was two years and “goodbye.” Not like today, when third and fourth years are possible . Then it really was two years and goodbye.
In the middle of the second year, I started looking around for a job because there it was, two years and goodbye. What I knew was atomic physics, so I looked around for an assistant professorship in atomic physics. Then one day Jan said, “How would you like to come to NBS?” This was in 1969. It was after I’d been there for almost two years. I said, “Well, let me think about that,” and then he made me an offer.
Levine:
- Audio File
- MP3 Audio
(Levine clip 2.mp3)
As I recall at the time, Ernie Ambler was the head of what is now called PML [Physical Measurement Laboratory]. It was called the “Institute for Basic Standards,” I think. I was made an offer, and I joined the Division, which doesn't exist anymore, called Radio Standards Physics, which was over here.
Kitching:
Over here at NIST in Boulder.
Levine:
Yes, it was the predecessor of what you see in [the Boulder Lab’s] Wing 5 now. That's what was there at the time. Don Jennings was the Section Chief and we had two big projects. The first big project was measuring the power of big shoot-em-up lasers with big calorimeters and that may still exist somewhere around here. The second project was what turned into measuring the speed of light, although it wasn't that originally. The idea of the measurement of the speed of light was that you wanted to measure the frequency of something and the wavelength of something and you're going to multiply them together, and out was going to come the speed of light. The ultimate goal was to make the speed of light a constant, and thereby link the standard of length to the standard of frequency.
That was sort of shimmering in the horizon. It didn't actually start yet. I started working on a number of laser projects. I built CO2 lasers, CO lasers, and HF lasers. Some of these lasers were for a project in molecular spectroscopy, I think. The HF laser was intended to be part of a lidar system and I was working with a contract from a company on this work. The HF laser was interesting because the laser action was in a flame, so it did not require a large external power supply. It could be flown on a satellite. Then about two years later, that Division was destroyed. It was split: half of the Division went to Gaithersburg and half of the Division was reorganized into the Time and Frequency Division. I woke up, along with a whole bunch of other people in the fifth wing, and we were now in the Time and Frequency Division. Jim Barnes was the Division Chief, but he really didn't know what to do with us because we were all “laser-ology” people of one sort or another. We were measuring the speed of light, or at least that's what Ken Evenson and his group were going to do.
The speed of light project then became serious. The goal was that Ken Evenson was going to measure the frequency of something and I, and Jan Hall and a guy named Dick Barger were going to measure the wavelength of the same thing. The thing that we chose was the 3.39 µm helium-neon transition, for which Jan had developed the stabilizer using methane.
Kitching:
Now at this time, the meter was defined in terms of the Krypton radiation.
Levine:
Right. The meter was defined in terms of a line in Krypton.
Kitching:
You had to compare your 3.39 µm laser with a Krypton lamp.
Levine:
Absolutely. That was the goal, that's right. The 3.39 µm wavelength was going to be compared against Krypton and at the same time, Evenson was going to make a chain in which he took some microwave frequency and multiplied it up and then locked some laser to it and then multiplied that laser up, locked on other laser to it, and step, step, step, step, step, step, step, and then you've got the 3.39 µm in, oh, I don't know, five or six or seven steps.
Each one of those steps required a Ph.D., so he had a large group of people who were measuring the frequency. The question was how are we going to measure the wavelength? The initial plan for measuring the wavelength was to build a 30 meter interferometer in the Poor Man Mine. This was going to be a 30 meter Fabry-Perot, which was in the Poor Man Mine because it was away from the city, it's up in Four Mile Canyon, it's free from vibration, and so on and so on and so on. JILA at the time was a pretty lousy place from the environmental perspective, from the vibration perspective.
The frequency project took a lot longer than anybody thought it was ever gonna take. Meanwhile the interferometer in the mine was just sitting there, so Jan said to me, “Well, if you want the interferometer, you know, just go play with it.” I said, okay, and I built a laser strain meter. The idea of the strain meter was I was going to have two lasers. The first one was a 3.39 µm laser, which was going to be locked to one of the fringes of the Fabry-Perot. As the cavity changes its length, the laser is locked to that frequency, and changes its frequency. Then there's a second laser, which has locked to Methane. I'm going to beat those two lasers together and take the difference frequency.
Kitching:
These are both HeNe lasers, one locked to the cavity on one locked to a methane transition.
Levine:
Right. They're both 3.39 µm, so you can't see them. You've got to line up a 30 meter interferometer cavity with a beam that you can't see. All right? Never mind how you do that. It turns out that it took me two weeks to do it the first time, and I was shaking. I couldn't stand it anymore because when you work in the Poor Man Mine for the whole day, you began to hear voices. You begin to hear sounds and you go a little crazy, right? You’re working in this tunnel.
Kitching:
Was there lighting in there?
Levine:
Yes, there were lights [and] there was power because there's all this locking electronics. I think you get the concept. There's one laser that's locked to the methane device, one laser that's locked to the cavity, you beat the two lasers together, you choose a fringe and the cavity such that the beat frequency is a Megahertz kind of frequency. The fridges of the cavity are five megahertz apart because that is c/2L, so you can always find a frequency that's within five Megahertz of the methane line. Then you record the beat frequency. In the early days, we had an analog version of the beat. We had essentially a frequency-to-voltage converter that converted the frequency to a voltage, and digitized the voltage and recorded it.
Initially we used paper tape. Later on we got fancy, but initially it was paper tape. As the earth moves, the cavity is pulled back and forth, so you can see earth tides and all sorts of stuff. I began to look at geophysics and that was my entry into the geophysics business. We began to look at earth tides, which are the equivalent of the ocean tides, except they're in the earth and you could calculate them and so on. There are a lot of early papers on the characteristics of the earth tides. One of the interesting results was a measurement of the nearly-diurnal resonance in the response of the Earth due to the interaction between the core and the mantle. The theoretical value for the frequency of the resonance is very close to one of the diurnal frequencies of the tidal potential, so that estimating the non-resonance portion of the response of the Earth turned out to be quite difficult. The analysis required the Fourier de-composition of time series that had several hundred thousand terms. A similar project involved trying to detect the response of the Earth to gravitational waves from pulsars or from rotating binary stars. The frequencies of the gravitational wave signals were known from astronomical data, so that these experiments also depended on Fourier analyses of very large data sets. We were able to detect strain-meter signals with amplitudes as low as 10-17, but this is not nearly adequately sensitive by several orders of magnitude.
Then we got into the nuclear explosion business because we're not that far from the Nevada test site. You could detect nuclear explosions. The interesting thing about the laser interferometer was it had a very wide bandwidth because the bandwidth is limited by all these electronics. Essentially, from the perspective of the seismic signal, the bandwidth is infinitely large, so you could see the first motion of the signal very clearly. A nuclear explosion and an earthquake are fundamentally different in the first motion because a nuclear explosion is a point source that just goes boom, and explodes in all directions. Whereas an earthquake is a sheer, it's a slip. [Therefore] you could tell the difference at least in principle, by looking at the first motion. We got into that business of could we tell the difference between nuclear explosions and earthquakes. The answer is yes you can, but the issue is how small a nuclear explosion can you distinguish from an earthquake. That was the issue at the time. Everybody knew you could distinguish it. However, the magnitude of an earthquake depends on the volume of material that ruptures during the event, so smaller events tend to look more and more like point sources. The question was how small you could distinguish it and the laser interferometer had some advantage in that respect.
Kitching:
This was at a time when there were a lot of underground nuclear tests? It was the height of the Cold War.
Levine:
Oh yes, we were plugged in to the schedule, so we knew when a test was going to happen. There were also tests to attempt to liberate natural gas by underground nuclear explosions that were intended to fracture the rock and liberate the gas that was trapped. A test would typically go off at sunrise or early in the morning, 6:30 in the morning, six o'clock in the morning. Since we knew when the test was going to happen, we turned up the bandwidth or recording system and so on and so on. There were a number of papers on the discrimination problem and it was semi-classified. A lot of the stuff was classified and a lot of stuff was not classified.
Kitching:
You were funded to do this?
Levine:
Yes. This was not an official affiliation, but yes, it was funded in a vague sort of way. However, Jim Barnes was not all that happy. He had a list and it was, “must do,” “should do,” and “could do.” He had his list and I was down at the bottom of the “could dos.” When you're down at the bottom of the could dos, you're not the Division Chief’s favorite person. On the other hand, I was not exactly his “unfavorite” person either. Yes, I was funded, but I was not the Division Chief’s favorite person.
Kitching:
At that time, Division Chiefs played a significant role in determining what research was done. Is that right?
Levine:
Well, not only that, Jim Barnes had an active research program. He was not just a passive Division Chief who sat in his office and did the Division Chief stuff. In fact, the basis of what's now commonly called the Allan Variance really came from Barnes. He played a very central role in what everybody now calls the Allan Variance. I don't know what the relative contributions between Jim Barnes and David Allan were, but Barnes made a very significant contribution.
Kitching:
It's always difficult when you have collaborations or many people involved in a research program, to identify contributions in some reasonable way.
Levine:
That's right. On the other hand, you know credit does not satisfy a sum rule. Just because I get credit doesn't mean you can't get credit.
Barnes was not all that happy with what I was doing. At the time, there was a thing called a Deputy Division Chief, which doesn't exist anymore. The Deputy Division Chief was Helmut Hellwig who came to me one day and said, “You know, you deal a lot with all of these crazy time series that are divergent and so on and so on. Why don't you take a look at the timescale because that's one of the problems with the timescale.” The timescale at the time was kind of a crude thing. It didn't really exist. I said, okay, and Barnes, Allan, and I built AT1. The basic design of the time scale came from Jim Barnes and David Allan. I thought about time series from the Fourier perspective, but they were thinking in the time domain.
Kitching:
AT1 was the first NIST timescale.
Levine:
It is still the basis of the current timescale. It's been massaged and improved, but the original assumptions are still in there.
Barnes, and Dave Allan, and I would sat in the cafeteria downstairs every day at three o’clock in the afternoon. The cafeteria downstairs doesn't exist anymore either. We’d talk about how to build a timescale and I was the guy who actually built it. I was the guy who actually did the design, but the ideas really came from everybody. Barnes and Dave Allen made very significant fundamental contributions to the ideas, but I did all the programming. Initially, the design was on paper tape.
Kitching:
Let me just take a step back. You have clocks at this time? You had cesium clocks?
Levine:
Yes, we were past the age of old quartz oscillators. We had HP 5061s and 5060s. We had early HP devices.
Kitching:
These were cesium beam clocks.
Levine:
Yes, cesium beams. We had a few rubidium [clocks], but they were all we had.
Kitching:
No masers at this time?
Levine:
Oh no, this was way before masers. I think masers existed in some form. Ramsey may have been building masers, but they weren't commercial devices. We built a timescale initially built on time difference measurements of commercial cesium devices.
We made one measurement every day at six o'clock in the morning and it was punched down on paper tape. Then I carried the paper tape upstairs to the computing center, which doesn't exist anymore either. It was at the end of wing five on the fourth floor. That whole end of the floor was the computing center. We would run the tape into the computing center and put out a microfilm showing the performance of every clock in the ensemble because that was the only reasonable output. In fact, I still have most of the microfiche of that era with a summary of the results. We would run the timescale, but it was really a retrospective timescale because we really didn't know what was going to happen at the time that it happened. At the time, we did synchronization with LORAN. We did common view with LORAN to the Naval Observatory.
We also did portable clock trips. We carried a clock to what was originally the BIH [Bureau International de l’Heure], which was in the Paris Observatory. This was before the BIPM was in charge at the time service. The time service was at the BIH downtown at the Observatory in Paris and later at the BIPM. We carried a rubidium [clock]. In fact, I did it a few times where you carry a rubidium [clock] from here to Paris. You have actually carry two, on an airplane. You would buy a seat for the clocks. You have to make it in twenty-four hours before the batteries were exhausted. I would stop at Dulles airport and I would meet the guy from the Naval Observatory. He would come out and we'd measure the time difference between my clocks and the Naval Observatory because in those days, you had to change [planes] at Dulles.
Kitching:
The clocks were operating?
Levine:
Oh, yes! You're carrying the time! Yes.
Kitching:
The clocks are operating on a battery in the airplane?!
Levine:
Yes! They’re heavy. Heavy! They weigh about forty pounds each and there’s two of them. We were really schlepping along because you're carrying the time, all the accounting, UTC-NIST or UTC-NBS, and you're carrying it to Paris!
Kitching:
This was a time comparison, not a frequency comparison?
Levine:
Yes. My guess is the closest was a microsecond or a few microseconds or something like that. That was the beginning and then over the years we made hardware upgrades, we made software upgrades, and we changed to a dual mixer system. The earliest dual mixer system was in about 1979 or 1980.
That was a dual mixer system that was designed in the Division and built by a company called ErbTec. The signals from the clocks at 5 MHz were mixed with an offset frequency of 5 MHz – 10 Hz, and the phase-difference measurements were made by a time-interval counter that measured the difference between the 10 Hz signals. The offset frequency was synthesized from one of the cesium devices. We changed to the Sam Stein Timing Solutions system later on. The hardware was completely different, but the basic measurement strategy was the same.
In some ways, the original kernel of AT1 carried forward. In fact, looking inside the current timescale you can see many of the fundamental assumptions of the [original] timescale. Really the things that Barnes and Dave Allan and I decided in the early seventies. The fundamental assumption of AT1 was (and still is) that the noise in the data was driven by the white frequency noise of the clocks. This assumption determines the time interval between measurements – it cannot be too short or else the contribution of the measurement noise will be too large and it cannot be too long because the variance of the frequencies of the member clocks can no longer be characterized as white frequency noise.
Kitching:
Getting back to the time comparison measurements, you would synchronize your clock here?
Levine:
You would take the rubidium [clock] and you measured the time difference between the rubidium [clock] and the local version of UTC. Then you carry it to Paris and then measure the time against their clock. Then you carry it back to Boulder and you measure what the time is, with respect to Boulder. Then you take that elapsed time, divide it by two, and that's the one-way delay. That gives you an estimate of UTC-NBS minus UTC. At the time, the BIH had clocks that were offset or they knew the offset with respect to UTC, but I don't really know exactly how that worked.
Kitching:
How many times did you do this exercise?
Levine:
I did it twice, but everybody did it a lot. There were many trips. Essentially, you didn't want to pay for the trip, so you did it if somebody was going somewhere anyway, carry a clock to the BIH. You had to get to the BIH in twenty-four hours and that was possible, but not trivial. You had a change in Dulles and it was an overnight flight to Charles de Gaulle airport. Monsieur Granveaud would pick you up at the airport and drive as fast as he could from Charles de Gaulle airport to the Observatory, which is on the left bank [of the Seine River]. It’s near Denfert-Rochereau. If you didn't make it within twenty-four hours, the batteries would die, the clock would stop, and the trip was wasted. Many people did the trips.
Kitching:
What was the goal of doing this?
Levine:
The purpose of UTC-NBS was to steer towards UTC and to realize UTC locally and that was, at the time, the best calibration system that we had. It was before the days of general relativity and that wasn't important in those days yet. That was the best we had. Common-view LORAN was not as good. It had systematic errors because of asymmetries in the paths between the timing laboratories and the transmitter; these delays also had an annual variation, and so on and so on. This was the best ground truth method that we had.
Kitching:
This was ultimately a way of synchronizing clocks worldwide: people bringing clocks to the BIH in Paris, contributing to their timescale, and then bringing them back?
Levine:
That's right. There were many different timescales. There was a thing called A1 and there was A3Mean. There were a lot of questions of linking them to astronomy because at the time the standard was astronomy, or was within the transition between astronomy and atomic time. Then in about 1988, the BIPM took over the job. That is the time service moved to the BIPM and then they began to get serious about international time, about what is now called TAI.
The algorithm was written by a guy named Bernard Guinot, who was kind of the father of TAI, and he wrote the timescale, which is called EAL.
Kitching:
When you refer to the timescale, I think what you're referring to is the way of combining the information from the different clocks and synthesizing the time out of that.
Levine:
Right. You have a bunch of clocks and yes, I've talked about timescales assuming everybody knows what a timescale is. You have a bunch of clocks and the characteristics of all the clocks, cesium [clocks] and later hydrogen masers, is that their parameters are stable, but they don't necessarily realize the SI frequency.
For example, if you take a 5061 or 5071 [cesium beam clock] or any of these devices, they have systematic frequency offsets which are much larger than the stabilities. They give you a frequency which is nice and stable, but it's not the SI frequency. If the goal is to measure the SI frequency or to realize the SI frequency, you have to discover the “personality” of that clock. A timescale algorithm has what you might call a “personality” of every clock in which you project off the systematical offsets. For example, if you have a clock that gains a second per day, exactly, then you look at its time and you algorithmically subtract a second a day from its time to get what you think of as the correct time.
A timescale is simply a more formal version of that idea, where every clock has its offset parameters. A cesium [clock] has typically a frequency offset and an uncertainty. That is, what is its frequency offset and how stable is that frequency offset. AT1 and most algorithms have that kind idea. In a hydrogen maser, in addition to a frequency offset, there's a frequency drift. You have a frequency offset and a stability and a frequency drift and a stability. Now you understand that the process is under-determined because you have all these parameters, but you only have one time difference measurement for each clock. The question of how you partition what you read in terms of systematic offsets, stochastic measurements, and so on and so on. That’s what the timescale algorithm is supposed to do. Every timescale algorithm has to solve that problem in some different way. The goal is to realize the SI frequency.
Kitching:
What was the NIST primary standard like at that time?
Levine:
Oh, that is a somewhat embarrassing discussion. There was a thing called NBS-4. It was used as a clock in the timescale by the time I came here.
Kitching:
This was the cesium beam clock?
Levine:
Yes. Let me put it another way. The difference between a commercial cesium clock and a primary frequency standard is that in a primary frequency standard, you have developed ways for evaluating all the systematic offsets that you could only learn about from the data in a commercial device. There are all these subsidiary experiments in which you estimate the systematic errors, the Doppler shifts, the pressure shifts, this shift, that shift, and then you remove them. That's what a primary frequency standard is. It's a cesium device in which you know what all systematic offsets are and you've corrected for them. What comes out in principle is the SI frequency. NBS-4 was one of these devices, but it was not the primary frequency standard anymore. Then there was NBS-5, which was being cannibalized when I arrived, and the successor was NBS-6, which was the primary frequency standard in that era.
Kitching:
This was the really long one.
Levine:
Yes. It filled up all of a room and it was like six or seven meters long. It had an uncertainty of a part in 1012 or a part in 1013.
Kitching:
At this time, basically, NIST was between standards in a sense.
Levine:
The problem was that NBS-6 was a very hard device to use because the essence of a primary frequency standard is you have to evaluate all these systematic offsets, and that was very difficult with NBS-6.
Kitching:
Why’s that?
Levine:
It wasn't designed very well with the idea of evaluating all these systematic offsets. It only ran like a few times a year, so most of the time, from that perspective, the time scale was running blind, in that you didn't have a primary frequency standard. The only information we had on the performance of the AT1 time scale came from the monthly reports (Circular T) issued by the BIPM. The report for each month was typically received by about the 20th day of the following month, so that the information was several weeks old by the time we received it. NBS-6 was so difficult to use; it was hard to use. Then in probably 1988 or 1989, we were going to build number seven. It was not going to be called NBS-7, it was going to be called NIST-7.
Then we had a rather long time in which NBS-6 didn't exist anymore, and NIST-7 didn't work. It didn't work, and it didn't work, and it didn't work, and it didn't work, and it didn't work, and it didn't work, and it didn't work, and it didn't work. If you like to think of it this way, we were the keeper of the flame without a flame. It didn't work, and it didn't work, and it didn't work, and it didn't work. Then Don Sullivan, who was the Division Chief, went to Paris for something or other and he went to the LPTF [Laboratoire Primaire de Temps et Frequence]. He saw, the demonstration of a fountain.
Kitching:
This was Andre Clairon.
Levine:
Exactly. He saw Clairon's fountain and, in fact, he was present at some kind of press conference in which Clairon unveiled his fountain. Now you understand that the whole LPTF is filled with smart guys. It's a building filled with smart guys.
Kitching:
Indeed, it is.
Levine:
Clairon is a smart guy, but he's got a lot of colleagues who are smart guys. Don Sullivan came home and said, we've got to have one of those. At the time, we didn't have anything. We had NIST-7 that wasn't working. Then there was a crash program to develop a fountain.
Kitching:
This was in the late 1990s.
Levine:
We developed the fountain after a while, but the fountain had some of the same characteristics of NBS-6, namely, you'd only run once in a while. The result of that is that the timescale depended on other things, effectively, and now we began to get masers.
Masers are at the same time a blessing and a curse. The blessing is that their short-term stability is super-duper wonderful. The timescale algorithm tends to weight clocks based on the short-term stability, so it tended to give heavy weights to the masers. On the other hand, the masers had frequency drift, and that frequency drift gets baked into this timescale. You've got to get rid of the frequency drift somehow or other. There was a big fight at the BIPM at the time in the Consultative Committee [on Time and Frequency] on whether we should allow masers into the timescale of TAI for the same reason. The EAL time scale algorithm that was used at the BIPM at that time did not include a frequency-drift parameter in the models of the contributing clocks.
Levine:
The problem was that there were a number of laboratories in which the primary ensemble was based on masers. Therefore, if you allowed masers in, you had the problem that you now have to deal with the drift of the masers. If you excluded masers, you now have to deal with the fact that you were excluding a certain laboratory, so you understand that there was a technical issue here and a political issue here. Eventually, we decided to include masers. Around the same time we changed the definition of the cesium to be the cesium frequency at zero Kelvin because there was a conflict between cold atom standards and hot atom standards because of the black body [radiation] shift.
The cold atom standards had different black body [radiation] shifts and you had to deal with that issue. The initial definition was based on thermal beam standards. When you introduce cold atom standards, you've got a black body shift of, I don't know, 2 in 1014 or something. Those were the two political hot issues of the day. We were running masers and we wanted a mixed ensemble with some masers and some cesium standards. We did a lot of work trying to estimate the drift of the masers. I would say we were only semi successful at that time.
Kitching:
You mentioned this issue of politics and the fact that time is ultimately a fundamentally international endeavor. What do you think of the ways in which politics, national pride, other things have coexisted with the technical work in this area over the years?
Levine:
The BIPM is a treaty organization and a treaty organization depends on the dues at the member states. It is technically outside of any one country. It sits in its place at Sevres, but that little piece of land is probably technically outside of France. We always have a problem with keeping the lights on at the BIPM and that means you got to get the countries to pay their dues. There've been a number of times when politics have intruded into things. You have to understand that you can't completely do what you would like, without any regard for the fact that the US State Department is paying some significant fraction of the budget of the BIPM. I don't know what the number is, but it's some significant fraction. Somebody who goes to the State Department and complains has to pay attention to that. You can't completely divorce yourself from it.
I have been a member of a number of committees of the BIPM for many years. Most of the discussions are about the realization of International Atomic Time (TAI) and Coordinated Universal Time (UTC). An important topic of current interest is the method that of adding leap seconds to UTC to maintain the magnitude of the difference of UT1-UTC to be less than 0.9 s.
Kitching:
I wanted to go back to the formation of the Time and Frequency Division at NIST. You were here, basically, when that happened. What was the Division like? How has it evolved over the years?
Levine:
When I first came to the Division, we had three very broad groups. There was what we called the “fourth floor,” and the fourth floor ran the services. Roger Beehler was in charge of the radio stations, there was Dick Davis, Marc Weiss, Jim Jesperson, Wayne Hanson. (I think Mike Lombardi came somewhat later.) There was a large group on the fourth floor that dealt with services of one sort or another. We had a proposed WWV-S satellite service. We had all sorts of things, so that was one group. Then there was the “second floor” and the second floor ran the timescale and the primary frequency standards such as they were, and then there was the fifth wing. The fifth wing was the people who were doing things Barnes would call the “could-do” things. [Ken] Evenson was doing molecular spectroscopy. There was a lot of stuff that was sort of in the could-do list. There were also a large number of people who were sophisticated technicians. I don't mean soldering guys. but people who were really sophisticated technicians. I think Dick Davis was probably the best of them. He basically built the early NBS GPS receiver.
Those were the three groups, and there was a lot of what you might call engineering support. For example, when we built the ACTS telephone service in 1980, I did the software and Dick Davis and Marc Weiss did the hardware and the hardware was built on Z-80’s, which was a microprocessor chip at the time. I don't think we could do that project now. Maybe we could, I don't know.
Kitching:
Because we don't have the technical expertise?
Levine:
Yes, we don't have the technical expertise and because we have replaced technicians, and I mean high level technicians, with other people who see life in terms of Physical Review Letters. I guess maybe that's the way to say it.
Levine:
That was sort of the Time and Frequency Division at the time. Once the speed of light was measured, most of the people in that fifth wing left or retired. Many people had been around a long time and, in fact, many of them have now died, some quite some time ago. In some ways the fifth wing, at least from my perspective, still has the same kind of flavor. It's still doing a certain amount of stuff that Barnes would put on the “could do” list rather than the “should do” list or a “must do” list.
Levine:
Do you understand that in some ways, the “must dos” have taken second place, maybe third place. For example, if you go upstairs across the street from the Division Office and you look at the pictures. Count the number of people who are detecting Schrodinger’s Cat. Schrodinger’s Cat is a fine thing to do, but count the number of people who are detecting Schrodinger’s Cat and compare them to the number of people who are running the radio station. I think that has happened slowly, but it has happened always in the same direction. That is, there are more and more people who are doing Schrodinger’s Cat, and I don't mean to pick on Schrodinger’s Cat particularly, but there are more and more people who are doing that kind of stuff and fewer and fewer people who are doing the intermediate stages between the research end and the services end.
Levine:
Let me give you an example of what I'm saying. If you go to other timing laboratories, many of the major time laboratories have optical frequency standards embedded in their timescale in some operational way. We don't. My private opinion is we're never going to have one. We're not going to have one in the foreseeable future because there's a big gap in the program between the guys who run the radio station and the people who develop the optical frequency standards in that the optical frequency standards have a high probability to remain laboratory curiosities. As long as I've been here, people are developing the next generation of optical standards. It was ion standards for a while, and it was optical frequency standards, but they have never gotten into the operational aspect of things because there's this big gap in the middle between the Physical Review Letter and making a device that actually works. You know, where there's a green light in the front and when the green light is on, it's working. That is the biggest change from the old days.
Kitching:
I wanted to ask you about the Internet Time Service that you were essentially developing. Can you talk about that?
Levine:
Let me tell you about the Internet Time Service. So that started out as a joke. Just like everything else. You see, in the later stages of my geophysics days, I had geophysical instruments in Yellowstone National Park to study the stability of the caldera. I had instruments in Southern California to look at the San Jacinto and San Andres Faults. I got into the business of running these remote systems. An important aspect of running the remote systems is knowing that they have accurate time tags. After all, if you're gonna measure the arrival time of an earthquake, you've got to know when the earthquake arrived. That meant that for my own purposes, I had to develop ways of transmitting the time to these remote sites. In those days, it was done with dial-up telephone lines because that was way before the internet. We had dial-up telephone lines to all these sites and I had like ten telephone lines to Yellowstone, and dial-up telephone lines to Erie, and dial-up telephone lines to the Poorman Mine. However, the Poorman Mine was sort of shut down at the time because it was too dominated by local effects. You had to develop this method of transmitting time across dial-up telephone lines. Then the internet came along and I said, “Well, what about that?”
Levine:
That wasn't gonna help Yellowstone and it wasn't going to help the San Andreas Fault because you don't have the internet out there. You only had dial-up lines. The first Internet Time Service, so to speak, was between the main NIST building and JILA and that had to do with exchanging GPS data because at the time GPS data was gotten by dial-up. You had to dial up all sorts of places and you needed a schedule essentially to dial up these places and you needed remote clocks and so on and so on. As I say, all of that stuff was developed for my own purposes or for purposes that really had nothing to do with the internet time service. Then I said, let's do it. Let's make it.
Levine:
I had one time server in a room at NIST.
Kitching:
This was a computer that was connected up to the timescale in some way?
Levine:
Initially it was dial up into ACTS [NIST’s Automated Computer Time Service]. It got its time by dialing into the ACTS telephone system, which we had also developed. And the ACTS telephone system was essentially running in a room upstairs. It got its time from ACTS, and it depended on the stability of the telephone lines. Then people started using it.
Kitching:
The internet time service?
Levine:
Yes! I had a few hundred requests a day. Then I got an NSF project for looking into the question of developing an internet time service.
Levine:
There had been a number of format definitions. There was a guy named David Mills at the University of Delaware and he had a definition for how you transfer time on the internet, the Network Time Protocol, or NTP, which at the time was just an emerging standard and now it's become the de facto standard. But because I had an NSF grant, I was allowed to put systems at NCAR [the National Center for Atmospheric Research] up on the Hill. I had eventually three, four computers up at NCAR and they were synchronized with dial-up telephone lines, that is, they dialed up down to NIST to ACTS. That's what made it go. That's what made it possible, was the idea of using the ACTS system and depending on the telephone lines for stability, because in those days the telephone lines were analog lines and they really were pretty stable.
Kitching:
What was the time reference for those systems? Was it a quartz oscillator?
Levine:
The local time reference was a quartz, but you synchronized the quartz with periodic dial up calls to ACTS. It resembled a timescale. You developed the personality of the quartz. Between calls, you've corrected the quartz oscillators for their frequency and the drift and so on and so on. A lot of it was carrying the ideas of a timescale into these remote sites. The first remote site was NCAR. This was run essentially without any internal [NIST] support. Zero. I had the NSF contract and NCAR allowed me to connect up there and the telephone calls cost nothing because they were local calls.
Levine:
Then I kept getting calls from people who wanted to host a time server. The deal that we worked out was that I would give them the hardware and they would provide the space and the power and the bandwidth, so it would cost almost nothing.
Kitching:
When you say bandwidth, this is bandwidth to the outside world, is that right? So it would allow more people to interrogate the time server.
Levine:
That is right.
Kitching:
That the main reason why your early server was up at NCAR?
Levine:
Yes, because NCAR had a major pipeline into the internet. NCAR was one of the main NSF computing facilities. It had a serious pipeline, much better than down at NIST, which at the time was terrible. We went through NOAA and it was terrible.
Kitching:
All this was in the early 1990s, I guess, right?
Levine:
Yes, in the 1990s. I began to get calls from a number of people who said, “If you give me the time server, I will provide a place of power and the network bandwidth so that it can be publicly visible and I won't charge you and you won't charge me.” I bought the hardware and I was funded by various agencies at the time who were prepared to allow me to buy hardware without asking exactly what I was doing with the hardware.
Levine:
The empire grew this way and eventually I had, like, 25 remote sites.
Kitching:
It sounds like NIST didn't fully appreciate the impact that you had achieved.
Levine:
I would have said it much more negatively than that, but yes, they did not appreciate it. They did not appreciate it. They did not support it. I guess it was more that they left me alone. There it was; I had twenty-five sites, several hundred thousand requests a second, and I was suddenly very visible.
Then, two problems came. Eventually I had more sites: I had service at NIST Gaithersburg which worked the same way. They had dial up access to NIST Boulder. Then two things happened that would not be so wonderful. The first thing that happened was that as the sites gained in popularity, the host locations realized that they had bought into something they weren't quite prepared for because remember they were providing the bandwidth for free. That was okay when it was not a lot of bandwidth, but after a while the tail was wagging the dog and these people would say “Well, ahem, ahem, we're not so much interested anymore.” That was the first thing that happened.
The second thing that happened was that the telephone system gradually changed over from pure analog physical lines to packet-switched digital networks. That meant that the stability of the telephone system was now very poor. The idea of using dial-up ACTS to control these servers was not all that wonderful. It wasn't going to work as well. It was clear that a crossroads was approaching.
We're into the 2000s. The crossroads didn't happen all at once. It happened slowly because one site would say “I'm sorry, we can't host it anymore. Here's your hardware back.” Of course, that would push load onto the other sites. The other sites would say “I'm sorry, we can't host it anymore. Here's your hardware back.” We push the load out to some other sites. Then at some other site, the telephone line would get bad and you couldn't use the telephone line anymore, and that would push load onto some other sites. You recognize that you were into a spiral that was not going to end well.
By now I had several hundred thousand requests a second, and the [NIST] management was very proud to stand up and announce how wonderful it all was and how with infinite foresight, they had seen all of this stuff coming and so on and so on and so on. You understand that credit does not satisfy a sum rule and that's fine.
I recognized that I was headed for a death spiral because the combination of the increasing load and the ACTS problems meant that what I had was unsustainable. In maybe 2010, the management realized that there were two problems. The first problem was the Judah understood what was going on, but nobody else had a foggiest idea of how this thing works because by now it was a major undertaking. This was not just some little desktop PC. There were twenty-five sites, there were dial-ups, there was maintenance, there was this and that and the other thing, and it was big time. Judah’s the only one who understands what's going on, and what's worse than that is, not only is there nobody else who knows, nobody else really wants to know. You know what I mean? Everybody else is busy publishing Physical Review Letters. However, they're not interested in this kind of stuff because it's not quite technician stuff.
You got to really keep your brains turned on. The project is somewhere between a routine technician job and a high-end research project, so we went through two modes. The first mode was: we're going to give it away, so for a number of years in my performance agreement, one of the deals was [Time and Frequency Division Chief] Don Sullivan said, “You’ve got to give it away.” The trouble with giving it away was that nobody wanted it, because we were giving the time for free and it costs real money to provide the service. You have a business model where you're giving something away for free and it costs you real money. In the modern era, there are companies that do that all the time, right? They essentially charge you for something else that you don't realize you're being charged for. It wasn't clear that there was any customer at the time.
Kitching:
These were big internet companies that you were talking to?
Levine:
Yes, they were. We were talking to them, but they weren't really talking to us. They said “Go away, kid, we're busy providing Google maps or something or other.” Maybe we weren’t talking to the right people, but it was hopeless. That was in my performance agreement year after year after year: you've got to give it away.
The other thing that happened at about the same time was the incredible rise of the computer security empire. Part of that was the result of security incidents, and part of that was the result of empires wanting to grow. They took one look at the computers on this internet time service, and they said, “Holy Moly, what are you doing? You don't have physical control, you don't have logical control. That stuff is out there. We just cannot allow this to go on.”
Kitching:
They were worried about it getting hacked?
Levine:
Not only that, it's at places where you only had the vaguest sort of idea of what they actually did. Some of the places, for example, I had time servers at Microsoft and Microsoft’s a serious place. I had to go through security audits at Microsoft. At one time, I had time servers at Microsoft headquarters, and then I had a number at Microsoft field sites, but I had servers that were operating in places that lacked adequate security documentation and controls. The security assessors say, “Well, this just can't go on. We don't have any oversight of what's going on. We don't know what's happening, blah, blah, blah, blah, yada, yada, yada.”
The next proposal was to rent space. We're going to go to these co-location facilities and install the time servers there. In fact, a whole bunch of people from NIST, Aaron Fein, and I went around to various co-location facilities and we were going to rent space in a co-location facility. That was what was going to happen, but you needed some kind of local time reference at a co-location facility so some sort of local reference time scale would also have been required.
At about the same time, this was in about 2011, the question was “What happens if the terrorists blow up the [NIST] Radio Building? What's going to happen to the timescale?”
The answer is we are going to build another time scale in Fort Collins. We chose Fort Collins because it is fifty miles away, and that satisfies some theorem or other that a backup site has to be at least fifty miles away. Judah is going to build a timescale in Fort Collins, which will have most of the aspects of the timescale at NIST, so that when a terrorist blows up the NIST building, Judah is perfectly fine and dandy even though he lives two miles from the NIST building and he's going to drive up to Fort Collins and it's going to be wonderful. Then the system said “We really don't want to use all these co-location facilities. What you want to do is you want to build internal fancy sites to support the Internet time service,” so we built internal fancy sites. There were essentially four of them: there’s WWV [in Fort Collins], there’s JILA, there's NIST-Boulder and NIST-Gaithersburg.
Kitching:
These are fancy sites where you have a time server and you have a large internet connection that pipes the time out?
Levine:
That's right. You have cesium standards at all these places. You have a timescale ensemble, and it's synchronized to Boulder somehow to the official timescale. It has enough local reference and enough local stability that essentially it could run forever. These sites have to handle all the traffic that twenty-five sites used to handle, so they have to have fancy network connections. Each one of these sites has a fancy network connection and that costs real money. I don't know the number exactly, but it's real money. This was regarded as wonderful in that it's now internal to NIST so I don't have to worry about the security issues and the security officers could be happy, or at least some of their unhappiness is addressed. That's where we are now.
Kitching:
Tell me about the impact of the Internet Time Service.
Levine:
Yes. I think two things have happened that we didn't really expect to have happened. Right from the very beginning, the Internet Time Service was designed to be independent of GPS. Initially it had ACTS and then later on it had these local cesium standards. That turned out to be a very important consideration in the modern era when people are concerned about jamming and spoofing of GPS. The fact that the Internet Time Service is independent of GPS, in fact it’s essentially stand-alone and independent of almost everything, has turned out to be an important advantage. That's the first thing. The second thing is that the financial and commercial places have increasingly serious requirements with respect to time accuracy and timestamping. Many of these places require a reference to a national timing laboratory, which in the US is NIST.
I have made two responses to that. The first is to offer a version of the Internet Time Service, which has digital signatures and that doesn't improve the accuracy of the service, but deals with the fact that the messages are not corrupted or not modified during transmission. More recently, we've talked about the idea of providing dedicated time service to some select places because the accuracy of the time service depends primarily on the stability of the network delay between the server and the user. The idea is if you had a better network, you'd get better accuracy. We have a thing called the NIST Special Calibration Service, which is intended to provide dedicated services to serious customers. One of the customers we've talked about all along is the stock exchange computing facility, which is in New Jersey.
The idea has been to run a dedicated, high class circuit between NIST Gaithersburg and New Jersey. That's shimmering on the horizon, and my guess is it's going to happen. Again, the advantage here is that this would be a time service that was completely independent of GPS. In fact, it has no external references at all. It's dependent only on the cesium [clocks] that were in Gaithersburg.
Kitching:
You mentioned the stock markets. What are the other application spaces that use the time service.
Levine:
You understand that we have 600,000 requests a second. When you have 600,000 requests a second, I don't know that I can characterize them, and I think there is sort of every imaginable use. The problem with Internet Time Service from the perspective of the super-duper serious users is that the stability of the network delay means that the accuracy of the time service is going to be a fraction of a millisecond under the best circumstances, let's say a few hundred microseconds for a good connection and a few milliseconds – maybe 10 milliseconds, for a more typical connection.
Levine:
That's not good enough for a number of applications. It's not good enough, for example, to support telecom. It's not good enough to support some aspects of power generation, but it might support other aspects of power generation. Most of these other things use GPS and the reason they use GPS is dthat it's there, and it's free. There's increasing concern about GPS vulnerability. That means that a time service based on fancy circuits is going to become more important. I expect the Internet Time Service is going to expand in that direction. There's always going to be a kind of vanilla service: every PC in the world and many network routers and switches synchronize with NTP. Then there's going to be a high-class service, which probably is going to cost money, we’re probably going have to charge for it. My guess is that's the growth, the growth area.
Kitching:
You were involved in an interesting legal case involving the use of timing in high speed trading. Can you describe that case?
Levine:
There have been a number of cases where time has played a role. The case where I was an expert witness was a case of insider trading. Insider trading means that you buy something based on knowledge that is not public yet. It's going to become public, but it's not public yet. You know something, and you act on it, and you buy a stock or sell a stock or buy a bond or whatever. Insider trading is fundamentally a time-based crime because what you're doing is not illegal, it's when you do it, that's illegal.
There was a guy who went to a treasury briefing and was told that something was going to happen and the price was going to change. He was told that was privileged information for another hour or something. However, he went out, called his ten best friends, and told them to buy it. The only problem with that is he got caught at it and that was a time-based crime. The question was, there were all these timestamps. There was the time stamp of when he leaves the voicemail message. Then there's a timestamp of when the trading company received the trade. The question is, what is the accuracy of all these timestamps. You would've thought of that as cut and dry, but it turned out to be not cut and dry. The primary problem was that there were no log files. That is, nobody had written down when the servers had been synchronized or tested or verified or whatever. J
It was not clear that you could prove it. If you could prove it, you wouldn't need me. I had to make essentially a hand-waving statistical argument that the accuracy was probably good enough because the accuracies of the times that were significant to the case were on the order of minutes, whereas the accuracies of the time servers were on the order of fractions of a second. That was a formal expert witness case.
Levine:
- Audio File
- MP3 Audio
(Levine clip 3.mp3)
There were a number of other cases in which I was an advisor, but not a formal witness. For example, I'll give you this, again a real case. There was a hold up at a 7-Eleven, and the 7-Eleven security camera had a picture of the holdup guy. The camera had a timestamp that showed what time the picture was taken and the photograph identified a person. However, the person claimed that this was a mistaken identity because he was somewhere else talking on his cell phone at that time and the cell phone records would prove it.
Levine:
The question is, what is the accuracy of the timestamp of the cell phone compared to the accuracy of the camera in the 7-Eleven? It turned out to be basically the same problem. The camera in the 7-Eleven had a clock in it, but nobody knew what the accuracy of the clock was or when it had been synchronized. I don't know what actually happened in that case. The problem there was that it was a crime in which time was an important issue because if the clock in the 7-Eleven was correct and his cell phone time was correct, he was innocent because he was really somewhere else. On the other hand, if the clock in the 7-Eleven was wrong, he could have held up the 7-Eleven. I don't know how that turned out, but the problem for that case was in fact the same problem. It was a lack of documentation for what the time actually was; a lack of log files.
Then there was another case. In this case, a driver driving a car hits a motorcyclist driving on the same road. I don't know exactly the details of it. The lawyer for the motorcyclist claimed that at the time of the accident, the driver of the car was talking on his cell phone. You know what time the guy was talking on his cell phone from the cell phone records, but the question is what was the time of the accident? How did that match with the time of the cell phone? There was an observer, a person standing on the side of the road or something. Somebody who saw the accident and recorded it on a camera or iPhone or something. However, nobody quite knew what the time accuracy of that thing was, so it was the same problem.
There are issues where time is important, but where there's no traceability, to use the technical term. At least in these cases, the accuracies of the times of interest are not anywhere close to the accuracy that the Internet time service can provide. The issue is traceability.
Then there are people periodically who call me up and say that they think eBay is cheating them because eBay has these time-based procurements. The goal I suppose is to transmit your bid as close to the possible as a deadline as possible. If you're over the deadline, you're excluded, so you can imagine the fight. How does eBay get its time? The answer is, I have no idea how. They do whatever they do and I don't know that it's right and don’t know that it's wrong. These issues come up over and over and over again. A thing that is common to all these issues is that accuracy at the nanosecond level is not the issue. It's traceability that's the issue. That is, we're not talking about nanoseconds here, we're talking about seconds, maybe even 10 seconds. However, being able to write down that yes, this is really traceable to NIST, that's a very common issue. I've had that issue over and over and over again and I don't know what else to say.
Kitching:
Interesting. One other thing I wanted to ask you about is that you worked at NIST for basically your entire career, but you're also a Fellow of JILA. What has been your role at JILA and how has that worked in terms of NIST?
Levine:
All of the NIST people who are Fellows of JILA are expected to have some active role in some department. In the early days of JILA, it was mostly the [University of Colorado] Physics Department. In the more modern eras, it's Physics and Chemistry and Molecular Biology. I have been a member of the Physics Department essentially as long as I've been in JILA. I've taught, not every year, but at least every other year I've taught something in the Physics Department. You don't get paid for it, but it becomes part of your official duty to teach. I've taught at one time or another, most undergraduate classes. I don't think I've ever taught a graduate class, but I've taught pretty much all the undergraduate classes at one time or another and, that takes a certain amount of time.
Kitching:
Do you enjoy it?
Levine:
Yes. I enjoy most of the teaching. You don't enjoy the “I didn't bring in the homework because my dog ate my homework” and all that kind of stuff. Or “I think I deserve a hundredth of a point more and can you change my grade.” This is the sort of stuff that you expect undergraduates to do. Yes, I find, if nothing else, it gives me a perspective on things that I didn't know about. For example, I've taught a course called Energy and the Environment which discusses the issues of energy balance, of where does energy come from and all that. That is very interesting. You learn a lot. For example, I had a guest lecturer who was the manager of the cogeneration facility at the university. Cogeneration is the idea of generating electricity and steam from the same plant. Essentially the waste heat from the electricity generator is then used to make steam. He talked about the technology of cogeneration and it was really very, very interesting. Another example, if you look at the SKIP bus, it says “high-efficiency diesel.” What exactly is “high-efficiency diesel”? What does it mean? I assigned that as a project to contact RTD and find out what it meant to be a high efficiency diesel. Some of the students actually found some engineering guy in RTD who explained to them how it works.
Levine:
That's another thing that's really interesting. In fact, they use a special fuel. It's not normal diesel fuel. Then there was the suggestion that if you took the waste heat from the refrigerator that makes the ice rink, you could use that waste heat to make hot water in the rec center. You freeze the ice and then of course, that has to reject the heat somewhere and say you reject the heat into the water and that makes hot water. And that's another interesting idea. Or you use the waste cooking oil as fuel, and you learn about these ideas and you get the students interested in these ideas. You find, it's a class of 75 or 100 students, but one or two of them are really gung-ho. When you say, “Find out how the cogeneration facility works,” boy they go find out how the cogeneration facility works It’s neat.
Levine:
I've also been chairman of JILA. I've had a role in committees of JILA, and that’ just a thing. It takes a lot of time.
Kitching:
We haven't really talked about your family at all. Can you describe your family and how that has influenced your career and how was the interplay between your family and your work?
Levine:
My wife is a book editor and we've been married for more than fifty years. And being a book editor has two incredible advantages and one of the main advantages is that you can work from home. She didn't always work from home; she worked at a place called Westview Press on 55th street. In the modern era, for the last twenty years or so, she’s worked from home. That means you can set your own hours to do your own thing. She has one room in the house, sort of her office. That has been incredibly useful. There are a lot of people who have to solve the two-body problem, you know what I mean? Especially two professional people have to solve a two-body problem. It's hard to solve a two-body problem. I don't know what the general solution to the two-body problem is, but the specific solution is hard.
Kitching:
You mean that one person has to compromise in terms of their careers?
Levine:
In some way or other, or everybody compromises. My wife and I had these complicated deals about who was going to stay home for the in service days of the school and, I'll pick up the kids at one o'clock and take them to music lessons. Since I took my daughters to music lessons, then you pick them up from music lessons. You have all these kinds of deals. My daughter studied violin in the Suzuki method, and this requires that the parent learn how to play at a basic level, so I learned how to play a quarter-sized violin. At that time, childcare was very difficult to get. I don't think it's all that easy now, but now you can do it for money. Then you couldn't even do it for money. We adopted two girls from birth essentially. My younger daughter died. She had a cerebral hemorrhage and eventually died. She was thirty-four, thirty-five, when she died. My older daughter lives in Longmont and she's not married, but she has a daughter who's five. I guess my youngest daughter’s death was very difficult. She was about thirty when she died, but she had been ill for many years. She had what's called an AVM, which is a brain hemorrhage, essentially. There were two possibilities with the brain hemorrhage. The first one is that you die right away. That's the end of that. The second one is that you don't die, but you don't ever get a hundred percent better anymore. You're kind of not so wonderful. She was not so wonderful for ten or fifteen years. She couldn't really hold down a job. She was essentially unemployed. There are a number of other people in the Time and Frequency Division who have similar arrangements. Helping a partially-disabled child is hard. It's hard to deal with.
Kitching:
How have you balanced your professional life with your personal life throughout your career? Has that been challenging for you? Is there any advice you’d like to give to you scientists facing these types of issues?
Levine:
Finding the right balance is difficult, and I don’t know a solution that would work for everyone. I have been very fortunate that my wife and I both have jobs that allow a certain amount of flexibility. This was especially important when my kids were growing up. Especially, in the early days, I did much less professional traveling than many of my colleagues, and I never brought work home in the evenings or went back to the lab after dinner. Even when the kids were little, and especially now that they are grown, my wife and I have taken an extended trip at least once a year. The trips are usually focused on history (a trip to Eastern Europe, for example) or on natural history (a trip to South Georgia Island or Jan Mayen Island). We have also been to China, Tibet, Thailand, and Vietnam as tourists – not connected to an official trip. We typically return with about 1000 pictures each, and we are very fortunate to be friendly with professional photographers who have been very generous with advice about Light Room and Photoshop.
Kitching:
You received the I. I. Rabi Award from the IEEE-UFFC in 2013. How did it feel to win this prestigious recognition?
Levine:
I was very honored to have my work recognized in this way. I have also received the Distinguished Service Award from PTTI, the Colorado Governor’s award for Research Impact in Information Technology, and a “trifecta” from NBS/NIST – a bronze medal, a silver medal, and two gold medals. I have also been awarded the title of “NIST Fellow,” a title given to only a small number of senior scientists. In addition to the very significant honor of receiving these awards, I know that the nomination and selection processes are lengthy and time-consuming, and I am very grateful to the people who were willing to spend time and effort to nominate me for these awards.
