Harry Sonnemann

Barth:

This is Kai-Henrik Barth speaking. It's about 5 minutes to 5:00 in the afternoon. I am sitting here with Harry Sonnemann in his home in Tyson's Corner, close to Washington, D.C.

Sonnemann:

That's right. Yeah.

Barth:

Mr. Sonnemann, why don't we start with a question about how you became interested in engineering on large projects?

Sonnemann:

Well, I actually was involved in underwater acoustics research program at Columbia University Hudson Laboratories and started there in the electronics department and gradually worked my way up to being a department head of the electronics department, and we were involved in seagoing experiments with regard to underwater acoustic program for the Office of Naval Research. And as we did the research, which was to a significant extent oriented towards looking at better ways to detect submarines and to determine what kind of characteristics the ocean has that would support better detection capabilities, we built sensors for measurements in the ocean [at] various depths. And in the late 1950s, early 1960s, we started to look at, the lab started to look at extended array configurations. And we built two strings of sensors and installed them off Bermuda to check to see whether there was coherence between the sensors and how well you could beam-steer an array. Which subsequently led to the decision by the laboratory and the Navy to build an underwater array that would essentially be the equivalent of a radar and try to, and also build a large acoustics source to ping on a submarine and then detect the returns and try to track the submarine satellite. So that's how I became involved in the large array business, so to speak.

Barth:

Your background, your technical training is…

Sonnemann:

Technical training was in electrical engineering at Brooklyn Polytech, and I have a B.S. in electrical engineering from there.

Barth:

When you mentioned the array techniques, which models were used for the array techniques in underwater detection? Were they models from the geophysical exploration side, or?

Sonnemann:

They were models basically from the underwater acoustics side. And the scientists that did most of the work at Hudson Laboratories the most significant amount that you may have to talk to, be willing to talk to is Al Berman. Have you talked to?

Barth:

No.

Sonnemann:

Well, Al Berman was the leading scientist that looked into the coherence of acoustic signatures, and we had acoustic sound sources we towed from New York to Bermuda and Bermuda to Nassau, and we received the signals on various instruments and then tried to determine to what extent the signals were coherent, and you could track the source that was sending out the signals. Most of the work was done at about 400 hertz, and that's what the large sound source also was that was used for the Artemis program.

Barth:

Could you describe the Artemis program a little bit more in detail? What was the major objective? How was it installed? Who were the, which organizations were involved in the installation?

Sonnemann:

Well, the Artemis program had two parts: one was the big acoustic transducer, which was handled by the Naval Research Lab, and the receiving part of it was done by Hudson Laboratories at Columbia University. And this was, that sound source is a 1-megawatt pulse source which was installed in a T2 tanker and then lowered below the water surface and then signals were received on the Artemis array which was installed off the Plantagenet Bank in Bermuda. And between 3,000-6,000 feet of water consisted of ten lines of 20 50-foot towers with floats on the top. And so there were 200 of these and each one of the towers had an array of sensors. And all those sensors were cabled back to [a] Texas Tower, and then signal processed and beam steered.

Barth:

When was Artemis operational?

Sonnemann:

That was done in the early '60s, when most of the work was done.

Barth:

So basically shortly, only a couple of years before LASA [Large Aperture Seismic Array] was designed.

Sonnemann:

Well, actually the work was mainly completed before LASA was done. Now, the director of the laboratory at Hudson Labs was Bob Frosch, and he was essentially instrumental in getting approval for doing the underwater installation of Artemis. The scientific community, the underwater acoustics community was not at all unified in that it made sense to do this because here was a great concern about the coherence of the ocean signals. And the proposal was made to do some tests with strings of sensors at various spacings, and check the coherence over time and find out what the signal-to-noise ratio is for and what kind of steering you could do. And it was certainly the view of Frosch that, and I agreed with that, that incremental implementation of this system would take maybe ten years and would be horribly expensive. So in order to make essentially an order of magnitude improvement in the state of the art, you had to install the whole array. And if you did that successfully and had the signal processor designed and built for that, then in 30 days you could verify that you have the signal-to-noise ratio gains [which] were [expected to be] in the order of 20, 25, 26 decibels, as designed. And if it didn't hang together, and the theory was just not valid, the whole thing would just go to pieces. So it was built and installed and the signal processor was put on the end of it, it was checked out, and it indeed achieved the signal-to-noise ratio gains that were predicted.

Barth:

And you were instrumental in designing this electronic device?

Sonnemann:

No, I did not do any of the design. I was essentially the engineering manager for the installation.

Barth:

But it was basically, the design was done at Hudson Lab?

Sonnemann:

The design was done at Hudson Labs with a large committee of scientific people from this community. And there were two factions. One was what's called the Artemis Committee, and then there was actually an anti-Artemis committee also.

Barth:

Who tried to shoot down the project?

Sonnemann:

Which didn't believe that this made any sense and that it was worthwhile to do.

Barth:

So, now to clarify this point, before underwater arrays there were just single instruments?

Sonnemann:

Well, there were single instruments and there were strings of instruments, strings of vertical instruments. So we did get some information on the coherence of signals and…

Barth:

But the array technique was really there tried for the first time on this?

Sonnemann:

Well, the array techniques, you can call a string of sensors an array.

Barth:

Is a vertical array, sure.

Sonnemann:

Is a vertical array. But then if you get gains that are not sufficient for detection purposes. So in order to get the gains you had to have an array that was big enough to get signal-to-noise ratios in the order of 20 to 26 dB.

Barth:

This is what basically came out for LASA in the end as well, 27 dB.

Sonnemann:

That's right. And so the dimensions for Artemis were roughly a mile by a mile.

Barth:

Because of the higher, uh, shorter wavelength.

Sonnemann:

Shorter wavelengths and the array itself was, you know, the height was 200 feet, so you had a 3-dimensional array. And we went through all the work associated with calibrating it and verifying that you know we could really locate.

Barth:

Was it at this point already something like multichannel processor to basically steer?

Sonnemann:

A very large multichannel processor. Because you had to take in, oh, essentially several hundred sensors.

Barth:

Do you remember how many sensors were involved?

Sonnemann:

No.

Barth:

Order of magnitude?

Sonnemann:

It's probably close to 10,000.

Barth:

Ten thousand hydrophones?

Sonnemann:

Well, I can look it up. I can at least check it. Let me see where that …

Barth:

I can switch off the …

Sonnemann:

So it's, I guess we said about 6,000 elements, and there are 32 in each, well there are 200 modules which were beam-steered, roughly, at an up angle, and then the final beam steering of all the 200 was done with the signal processor, which was designed by IBM. And the Artemis experience that we could in fact get the signal-to-noise ratios that were predicted were essentially the basis for the subsequent decision at LASA you ought to be able to do the same kind of thing.

Barth:

So Frosch was the …

Sonnemann:

So Frosch went from being the director of the Columbia University Hudson Laboratories, he went to the Advanced Research Projects Agency and became the director of the Nuclear Test Detection Office.

Barth:

And he brought with him managers like yourself?

Sonnemann:

He didn't bring, well he didn't bring any with him, but when the decision was made by ARPA to pursue the program I came down to Washington to manage it as the director of field engineering, assistant director field engineering in that office. And so I was again the person that was mostly concerned with making sure that the design was properly translated into engineering that you could install, because the problem was one of how do you do this kind of a job, because now this was 150 kilometers by 150 kilometers, with 525 boreholes [and sensors at a depth of] 180 feet. And then you had to wire them all up. The 25 sensors had to be wired into a sub-array, and then the 21 sub-arrays had to be wired into something for information transmitted. Now, so the design was done essentially by MIT, Lincoln Laboratories and Paul Green. Paul Green. Have you talked to Paul?

Barth:

No.

Sonnemann:

Well, he was, that group was instrumental in the design. And the decision was made to use log-period spiral so that you get varying distances between sensors, since we already concluded with the Artemis array that precise location wasn't important as long as you could locate it subsequently.

Barth:

As long as the distances was…

Sonnemann:

The distances were appropriate for it. And it was the other I think experience that came from Artemis is that Artemis was a very sparsely populated array, and that was okay too, and that you know the signal, the noise, that you could process out the noise and get the signal-to-noise ratio. So LASA was also a very sparsely populated array. So from a theoretical standpoint I don't think that there was problem, at least in Frosch’s mind, to translate the technology. But to convince the seismologists that that technology which was a factor of 100 different in frequency would be applicable to earth sciences which is different kind of background noise situation and reflections and refractions, then you have an order except it turns out in water you also have reflections, refractions and layers and it didn't interfere with the ability to process. But that was a major area of disagreement, and a major stumbling block you might say in acceptance of the idea that you should build a LASA. So you really have a situation where you had to have somebody in the right place, like Frosch in Nuclear Test Detection that had the, first of all, technical understanding, secondly the respect of the scientific community, and that could convince the director of the Advanced Research Projects Agency and the rest of the Defense Department that to invest $10 million in that array was a reasonable thing to do. I think one thing that was driving the decision that at that particular point there was great interest in improving nuclear test detection capability, particularly with respect to monitoring Semipalatinsk. And that kind of gain would make it possible if it was successful, if the array was successful, to detect nuclear events at Semipalatinsk down to a magnitude of 4.

Barth:

Which would be order of magnitude, I mean, depends of course on the geological structures between Semipalatinsk and Montana, [it depends on] the travel paths of the seismic waves what kind of yield this would be. The yield-magnitude relationship was probably something, we would have to check this, but I think it's probably something like 4.0 for this area.

Sonnemann:

Yeah, something like that, yeah. So that's why I said in my e-mail that I don't think that if we had had the Artemis experience on our belt, under our belt, Frosch had not been there and had not gone to ARPA, probably the Air Force would have continued to try to take the Tonto Forest size of array and gradually try to do more of it. Now when I came to ARPA, the question was that, "How do you do this?" I mean the design was done. The question is how do you install it. And the Air Force’s proposal was to build 21 Tonto Forest manned array sites.

Barth:

Very expensive.

Sonnemann:

And you look at that, not just expensive, but you now have to record and at the time you had only tape recorders, a couple 2-inch wide tapes running at reasonably high speed. Now you had to record at 21 sites, then take all these tapes, bring them to a central data center, build a signal processor that could, well, first of all you had to take all those 21 sets of tapes and play them simultaneously preferably into a signal processor to do whatever you want to do, beam steering to signal processing. And I just rejected that out of hand. I told Frosch that it didn't make any sense to me, that I thought the only way that we could do it is essentially to use what we did in Artemis. We had from the array sites we had cable going up to the signal processor, which happened to be on Plantagenet Bank, but that was a matter of convenience, so that what made sense is to take the 25 sensor sub-arrays, bring all the signals to a central vault, digitize them, and then send them over phone lines or whatever to a central processing center.

Barth:

But in this way the information of a single seismometer was not lost. So it's not that you just…

Sonnemann:

The information of — no, yeah.

Barth:

So there was already a summation made for the…

Sonnemann:

No, no. The information was not lost.

Barth:

Okay. It was just multiplexed, and everything.

Sonnemann:

Multiplexed. So and that required, well first it required something like a thousand miles of cable to be trenched in. Because the other thing I said, from the sensors to the central vault of these 25 sensors, if you had them all on the telephone poles it would cause a major problem because you couldn't use the surface and we would have to get rights-of-way and easements for having poles stuck up all over the place and…

Barth:

Very vulnerable too.

Sonnemann:

Vulnerable too to damage. So the most sensible thing to do, since the technology was there, is to bury the cable 3 feet belong the ground. So that's what was done. You know, you had 180-foot boreholes and you had a vault at the top, a little vault, and then buried cable to the center, so you have these central digitizers and multiplex it into a telephone system and sent it to Billings, Montana, to a signal processing center.

Barth:

And from there off-line things were directed to MIT?

Sonnemann:

Well from there, no.

Barth:

Lincoln Labs?

Sonnemann:

The signal processing was in Billings, and that signal processor was built by IBM by the same group that built the Artemis signal processor, so the technology was already there and they felt very comfortable that they could build one. And it was built, and again within a few weeks after the array was turned on it was pretty clear that you were able to get the signal-to-noise ratios that were anticipated. You could do better with some optimum processing than just straight delay and sum, but at any rate it verified that you got the gains that were anticipated. And the recordings were then made at that data center and you could then distribute off-line the information to other facilities for special processing. So most of it went to Lincoln Laboratories for processing. But the Air Force processed some; AFTAC [Air Force Technical Applications Center] processed some. The other thing of course they discovered is that they could also detect virtually all of the major earthquakes anywhere down to a magnitude of 4 or thereabouts. And that resulted in this earthquake bulletin that put out on a regular basis.

Barth:

To what extent was LASA used by the academic community?

Sonnemann:

Well, it was used by the, when the bulletin came out, that of course then allowed all the other seismological observatories to have a much better database to work with. But so I think it was a big help. And also the information from the other seismological observatories then was integrated with LASA so that you could verify that some earthquakes we detected in some location where there was an observatory agreed with the information that came out of LASA. So and from the LASA experience then you had the development of NORSAR [Norwegian Seismic Array], which was the array in Norway that was a smaller one.

Barth:

So in the end the hope was to have basically LASAs all over the globe and have large arrays.

Sonnemann:

There was one short period of time as I recall of not more than a week when the Secretary of Defense Office wanted to know how many LASAs it would take in order to monitor nuclear explosions globally, and we went back to them and said 11. And they wanted to know what the cost would be, and we said $110 million. Ten million dollars apiece for 11. And as I recall, they actually allocated the money for it for three days or four days, then decided maybe we should just think about that again. [laughs] But at one point we were actually looking at what it would take to globally implement it. Now with a multitude of LASAs, if they were detected, if there was an event detected at two of them and we had some reasonable signature information, then you could probably get another incremental gain in signal-to-noise ratio because now you know what the signal's characteristics are and you could do some optimum processing. So there was some hope that if you had a significant number of them that you could down to maybe a tenth of a kiloton, some low number.

Barth:

So the detection threshold with LASA came down to about, unified magnitude scale of 4?

Sonnemann:

Probably 4, a little over 4. You'll probably get a better number from Carl [Romney], because he followed it. I left in '68 which was a year after LASA was installed and went to the Navy Department.

Barth:

So you left already in '66 or?

Sonnemann:

I was in ARPA from '64 to '68, which spanned the installation of LASA and also the planning for NORSAR.

Barth:

Okay. But LASA was operational already in June '65.

Sonnemann:

Oh yes. LASA was operational.

Barth:

And what happened in the three years from '65 to '68? You mentioned that there was also construction going on with LASA?

Sonnemann:

No. No. There was no construction going on. I mean it was, once it was installed it was done.

Barth:

It was done.

Sonnemann:

I guess one comment that's worth while making is that one of the other things that was driving installing it the way it was installed and also with the underground cable source that, you were talking about 525 locations that had to be maintained, and I couldn't see any implementation that would make it possible to have reasonably consistent data and continuity of operation if we didn't have pretty high assurances that the breakdown rate and the repair rate was low. And also what we did is make provisions that from the central location you could essentially check out all this remote sensing equipment and see whether it works. And we could fly in a helicopter to a site that was not operational, substitute an amplifier, a seismometer, whatever needed to be fixed, and right there get verification that it was fixed. So in essence you could maintain the system with a handful of people and some reasonable decent transportation.

Barth:

Not like the AFTAC proposal having 21 manned Tonto Forest stations.

Sonnemann:

That was very expensive, but they could have maintained them there, but I think the signal processing would have been an absolute nightmare, because I don't know how you synchronize here. You could synchronize it, but all these tapes, and you were talking about a truck full of tapes a day.

Barth:

So the central piece of LASA is basically then the multichannel processor?

Sonnemann:

Yeah.

Barth:

Which can really deal with the data streams coming from the sub-arrays.

Sonnemann:

In real time.

Barth:

In real time. That was the most important thing.

Sonnemann:

That's right, yeah.

Barth:

I see that similar developments happened in the history of physics. I mean you have bubble chamber work to detect nuclear particles and at one point or another you have millions of photographs per year, so you have to figure out how can you just digest this amount of data, and so you had to come to some form of automation, and this seems to be a similar case, that just the amount of data produced every day had to be, had to lead to some kind of automatic readout or online processing; otherwise one would just drown

Sonnemann:

Well that's right. I think the experience from taking data almost in any discipline is that if you had paper recordings 95 percent of them wind up in a storage space and very few of the material gets analyzed. So we felt you really had to have essentially a beam formed that looks at the target in real time. And then of course you could also, in the case of Artemis you could steer it and follow the target. And we demonstrated that you could do that.

Barth:

I guess Artemis was highly classified and not accessible for academic research.

Sonnemann:

At that time it was classified, because it was essentially anti-submarine warfare program.

Barth:

What about LASA? I mean LASA was…

Sonnemann:

One of the concerns was that if you classified, then it will be extremely difficult to handle the data for one thing, but also that at the time I think the United States wanted the Russians to know what our capability was to detect them as a deterrent. So the international community was totally involved in this. And since there was great interest in monitoring nuclear events in Semipalatinsk particularly, we had the cooperation of the Norwegians, the Swedes, the Danes, the British, the Australians, and eventually the Japanese, and other countries that were interested in it. And there was good collaboration, and they all participated in it. And of course the Norwegians and the Scandinavians were very much concerned, and the notion of putting an array in Norway that was close enough to Semipalatinsk so you would get a larger signal, was interesting to all kinds of Europeans. But they also could make it smaller, because they didn't have all that real estate.

Barth:

It was probably a very sensitive negotiation process with the Scandinavians who might be very concerned that the Russians will take this as an offense.

Sonnemann:

Well, I think the Swedes would have liked to put it in Sweden except that Sweden had declared its total neutrality and I don't think they felt they could put it there. But they had access to the information if it was generated in Norway, and they could tie it into the Swedish array so that they can compare data, so I think that…

Barth:

Maybe we can come back in a little bit more detail to your actual day as a manager during these times. So what would you actually manage during this time? Would you be in charge to have communication with the contractors who would take data?

Sonnemann:

Well, I was essentially the person in ARPA that had the responsibility to make sure that the program got executed properly. The Air Force was responsible for the actual installation work. I had a lot to do with the, in essence directing the engineering implementation like, I guess I could say at the time that it's unacceptable to build 21 Tonto-like arrays, and so I was intimately involved in the decisions on how the design was to be done, not the design, how the implementation of the design was to be done in Montana. And that went as far as getting into some heated discussions with the Air Force about whether you have to worry about lightning protection.

Barth:

Well you have.

Sonnemann:

Pardon?

Barth:

You have to worry about this, because it's Montana and you have storms I guess.

Sonnemann:

Well, you look at this 150-kilometer grid and all the lightning that you have, it's certainly something that you should worry about. And but there was opposition from the Air Force to install lightning protection.

Barth:

That's interesting, because I mean of course Wichita Mountains and all of other seismological observatories had also lightning protections.

Sonnemann:

Yeah. So I insisted that they provide the space for lightning protection in the amplifying system. Because all these amplifiers were connected to a cable several miles long, and we, one of the arrays was built with lightning protection in it, so we had one sub-array of 25 sensors had lightning protection and the other 500 didn't. The first major lightning storm that came out took out 500 amplifiers.

Barth:

All just burned.

Sonnemann:

So.

Barth:

Well, I mean the pre-amps; they are so sensitive that it doesn't take much to kill them.

Sonnemann:

The other 25 survived, so they installed lightning protection in the other 500. So I had, at least I had enough authority so I could insist that they put it in one and provide the capability to add the diodes, to do the lightning protection and…

Barth:

So no fuses, just diodes.

Sonnemann:

Well, there were fuses, there may have been fuses and diodes, but at any rate then they installed them in all of them. After that they didn't have too many problems.

Barth:

How was the work of the LASA study group? I mean there were people in there like Frosch was the chairman, you were part of this group, and there was Carl Romney, there [was] Captain Meek as far as I remember.

Sonnemann:

Yeah. As a matter of fact, if you want to know, I mean I have, we have coordination meetings that had like Paul Green and the Air Force people of all kinds and AFTAC personnel, TI [Texas Instruments] personnel, Geotech personnel, and the Lincoln Lab people, and essentially that was on the engineering end and the installation end, and then you had orientation meetings for various people. These were all on the working level. And then we also of course had, let's see where we are. I'm trying to find…

Barth:

I can stop the tape for a second.

Sonnemann:

Yeah. [tape turned off, then never turned back on; end of taped interview]