Interview with Sander Weinreb

Sander Weinreb, 1936- . Interviewed 31 July 1978 at the URSI Meeting in Helsinki, length of interview: 50 minutes.

Papers of Woodruff T. Sullivan III

NRAO/AUI/NSF

Oral History

Sullivan, Woodruff T., III

Weinreb, Sander

Audio cassette tape

50 minutes

Late 1950s and early 1960s at MIT, Green Bank, and Lincoln Lab working on digitial autocorrelator, search for deuterium line, search for Zeeman splitting, OH absorption discovery; details of receivers and analyzers.

1978-07-31

The interview listed below was either transcribed as part of Sullivan's research for his book, Cosmic Noise: A History of Early Radio Astronomy (Cambridge University Press, 2009) or was transcribed in the NRAO Archives by Sierra Smith in 2012-2013. The transcription may have been read and edited for clarity by Sullivan, and may have also been read and edited by the interviewee. Any notes added in the reading/editing process by Sullivan, the interviewee, or others who read the transcript have been included in brackets. If the interview was transcribed for Sullivan, the original typescript of the interview is available in the NRAO Archives. Sullivan's notes about each interview are available on the individual interviewee's Web page. During processing, full names of institutions and people were added in brackets and if especially long the interview was split into parts reflecting the sides of the original audio cassette tapes. We are grateful for the 2011 Herbert C. Pollock Award from Dudley Observatory which funded digitization of the original cassette tapes, and for a 2012 grant from American Institute of Physics, Center for the History of Physics, which funded the work of posting these interviews to the Web.

Please bear in mind that: 1) This material is a transcript of the spoken word rather than a literary product; 2) An interview must be read with the awareness that different people's memories about an event will often differ, and that memories can change with time for many reasons including subsequent experiences, interactions with others, and one's feelings about an event.

Working Files Series

Individuals Unit

Transcribed by Sierra Smith.

Sullivan

Ok, this is talking with Sandy Weinreb at URSI [International Union of Radio Science] meeting in Helsinki on 31 July 1978 and as I said before, we are just going to cover the first part of your career but can you tell me what was your educational background and how you first in contact with radio astronomy?

Weinreb

Well, I was an undergraduate at MIT and had been always interested in receivers. That is, my boyhood ambition sort of was to build a very vast communication receiver and I worked a lot on that. I guess the summer of my junior year- I started in ’54, so this must have been around summer of ’57- I decided to spend the spring vacation going through the yellow pages of the Boston phone book and I just called up everybody under electronics and under physicist and I got about ten interviews and one of them was with the Ewen Knight Corporation. It’s in Needham and I interviewed with Pete [Shrum/Strum?] who was sort of the chief engineer and he was vice president of the company. Jack Campbell worked there and Frank Drake was a graduate student that had a part time job at Ewen Knight. And one thing I remember about the interview is that [Harold "Doc" Irving] Ewen was sitting on top of a large home built work bench. It sort of had a platform in the middle and he was sitting up there talking to people who were working around the bench and [Shrum/Strum?] had showed me around and he introduced me to Ewen and Ewen said something like, "Give him a buck. Put him to work." That was about it. They did offer me a job. Actually they didn’t offer me a job. They got Harvard to offer me a job through Tom Gold [Thomas Gold].

Sullivan

Who was director then.

Weinreb

Of the radio astronomy, yes, and the job was to make a hydrogen line receiver, 100 channel receiver, that did not work, that is to make it work.

Sullivan

100 channels?

Weinreb

Yes. There was a receiver commissioned, probably 1954 or so, Ewen Knight was to build Harvard a 100 channel receiver with digital output. It was supposed to type it all out on an electric typewriter and it was one of the first disasters in equipment in radio astronomy. It just never worked right.

Sullivan

What were the basic problems?

Weinreb

Well, it was unstable.

Sullivan

Was it DC?

Weinreb

I think a few different schemes were tried. I think that it was to be a DC comparison radiometer, in terms of there is a reference band which is compared continuously with each channel but I think there was trouble with getting square law detectors that weren’t too temperature sensitive. I think it had a lot of crystal filters in it. By the time I got to it, it was a very sore area sort of between Harvard and Ewen Knight. I guess Ewen Knight got paid for it but I don’t think it was ever used on the telescope.

Sullivan

Did Ewen Knight do other things besides the radio astronomy by that time?

Weinreb

Yes, they were building equipment for the military also, receivers for the military. I guess I started by trying to build a RS switch for 21 cm and I did that. But then we built a very modified receiver. That is, from 100 channels, we went to 1 channel and another student, Ron Weimer who now works at Green Bank, he came to work also for Harvard but at Ewen Knight and we put together a simple receiver. We built the filters, just LC filters, and we installed it at Agassiz Station. [T. Kochu] Menon and [Frank D.] Drake helped with it.

Sullivan

This is all the summer of ’57 now you are talking about?

Weinreb

No, it went on. I guess I worked part time during ’57, ’58, and it was around the spring of ’58 that we installed this receiver and I think I got out of radio astronomy the next summer. I went to work for Radiation Incorporated down in Florida. I worked on communication receivers.

Sullivan

What was going to be the advantage of this single channel thing?

Weinreb

It worked.

Sullivan

That it worked? They had nothing that worked? They had been taking observations for a year or two, hadn’t they by that time? But maybe not all that successfully.

Weinreb

They had something that worked but I don’t remember what it was, whether this was higher resolution or just where it fit in.

Sullivan

In any case it did stay on and was used?

Weinreb

I believe so, yeah.

Sullivan

So when did you next some into contact with radio astronomy?

Weinreb

Well, I started looking for a thesis.

Sullivan

You stayed on at MIT as a graduate student.

Weinreb

Yes, right. Let’s see, I graduated ’58. The first year I was a full time student. I had a fellowship and I did a lot of looking into communication theory. That and microwaves were sort of the two fields that I specialized in. And I should say I did an undergraduate thesis at MIT which I looked into the feasibility of detecting the deuterium line.

Sullivan

I see.

Weinreb

It was under Professor [William] Siebert at MIT who was a communication theory person.

Sullivan

Now the Russians had claimed it I think going back to, was that the Jodrell Bank meeting in ’55 I think they made a claim?

Weinreb

I don’t remember. I think at that time there had been two unsuccessful attempts: [?].

Sullivan

Right, in the west.

Weinreb

And the Russian claiming detection but for this undergraduate thesis I just wrote down what temperature to expect in terms of the deuterium to hydrogen ratio and just got an idea of the problem. I didn’t want to look for it at that time. I didn’t plan to continue.

Sullivan

Were there any suggestions in the thesis of what was needed electronically or to better look for it?

Weinreb

No, it just pointed out that the ratio of what you wanted to detect to the system noise. Well, it could either be an absorption experiment in which case that ratio was limited by the fractional absorption, something like 1 to 5000 or an emission experiment would require a tremendously low noise receiver so it looked pretty hopeless. Well, then I started looking for a doctoral thesis a year or two after that. And by the way, the suggestion of looking for deuterium came from Ewen when I worked there. He thought that would be a good thesis. Then the doctoral thesis I talked to several professors. [Jerome "Jerry"] Wiesner was one of them. Wiesner was interested in radio astronomy and he thought the deuterium line would be a good thesis. I started looking into ways of doing it, coming to the conclusion that stability and long integration time were the only hope, a multichannel receiver of course, and stability led me to think of digital processing. Autocorrelation techniques were a big thing at MIT in those days; that is Y.W. Lee and Wiesner and some other people had written papers about correlation techniques. So my original intent was to build a digital correlator with many bits. That is I didn’t realize at first that very coarse quantization could be used.

Sullivan

That it could be used without too much degradation?

Weinreb

Yes, I didn’t think that was possible but one of the first problems I started working on was how many bits did I need and the simplest problem was to take the case of one bit to see what would be the loss for one bit, never thinking that I would use one bit. And I worked that out sort of using statistics that I had, a signal processing theory, and I was sort of very shocked that you could make an unbiased estimate. That is that you could correct the mean of your result to get to the true autocorrelation function. I wasn’t even looking for that. I thought there was some error that no correction was possible and I don’t think I computed the loss in signal to noise at that time. I worked out this result so this sin relationship between the two normalized autocorrelation functions. The very first person I told it to was another graduate student named Irv Steglitz and Irv said, "Oh yeah, [van Vleck?] did that in 1943," and since I never told anybody after that who knew of [van Vleck?] paper but this first guy I met over in the drug store across the street from MIT and he had known about it.

Sullivan

He gave you that big clue?

Weinreb

Yes. It was...

End of Tape 108A

Sullivan Tape 108B

Sullivan

Ok, continuing with Sandy Weinreb on 31 July ’78.

Weinreb

Well, the next step was to compute the deterioration in signal to noise due to one bit quantization.

Sullivan

(van Vleck?) hadn’t worked that out?

Weinreb

I don’t think so. No, I don’t think he did. Of course, [van Vleck?] worked it out before there was digital processing. That is he didn’t think of it in terms of a one bit quantized signal. It was just if you had hard limited a signal, what happens to it. There was also some work in the Harvard Acoustics Lab, [Curran?] I think, [Curran?] and Hills, [Ferran?] and Hills, that had looked into hard limiting of Gaussian noise signals.

Sullivan

Does this mean clipping essentially [???]?

Weinreb

Yes, yes. So I worked on the increase in the variance of the signal and that was fairly easy to compute for the autocorrelation function but I was really interested in the variance of the spectrum which requires knowing the covariance of the autocorrelation function and that always eluded me. That is I think until just a few years ago that was an unsolved as far as an exact solution is concerned. I found it through a computer simulation.

Sullivan

A Monte Carlo sort of?

Weinreb

Yeah.

Sullivan

I see. So that was just an empirical number until just a few years ago?

Weinreb

Well, the increase in the autocorrelation function is pi/2, the increase in the variance of the autocorrelation function, the increase of the square root of the variance the rms deviation is pi/2 and that you sort of expected that to be true of the spectrum. You could prove that on the average the spectrum increase would be pi/2. Well, I guess I told Wiesner about this one bit method and that looked like a good thesis to apply that to looking for deuterium line. I planned out a receiver and I made a cost estimate which I believe came to a $60,000 NSF [the National Science Foundation] grant which Wiesner applied for. Some was overhead and so forth. I think it left about $30,000 for equipment. Some of it was my salary for a year and a half. So we applied for that grant and I forgot who the NSF guy was, Jeff something- he was quite good. We got the grant. Prior to getting it or maybe it was after getting it, we went to Lincoln Lab to try to interest them in building the digital correlator. That is Wiesner and I went out there and I remember there was an estimate of $200,000 to build it. That is it was going to require 1,000 flip-flops and at $200 a flip-flop was $200,000 which they did have that kind of money to just throw into something like that and I had budgeted $20,000 and in fact, did build it for that. It was built by a contract to a company Control Equipment Incorporation. It was a little tiny company run by somebody else who used to work for Ewen Knight, Harry Adams was his name, and for $19,000 they built the timing and correlation part of the correlator.

Sullivan

To your specs?

Weinreb

Yes. It did not include the counters that accumulated the autocorrelation function. This was a 21 channel machine and the counters, well, there were 21 of them and they had to be able to accumulate for several hours at 300 KHz and then manually read out and I think there was something like 20 flip-flops in each one so that about 400 flip-flops. It was a large part of the hardware and those I designed myself and there was a husband and wife that wired them. They did them at home at night while watching television. And these counters had neon bulb readouts and also an electromechanical counter that did the slow counting. Well, the correlator finally got built.

Sullivan

How were you doing the transform now?

Weinreb

The idea was to write down the numbers out of the correlator every four or five hours and then punch these on cards and transform them. At Green Bank there was an IBM 1620 computer that did the transform.

Sullivan

So you would have four or five hours of integration and then you would just write down the number?

Weinreb

Yeah. The biggest problem was a few of the electromechanical counters miscounted and the correlation function had to be extremely accurate. I mean just missing one out of a few counts with an electromechanical counter, which was some valid numbers, that was terrible but this only happened on one of these. What I did was run the machine for days, all the counters should count the same and I would run it for a week even and would find one electromechanical counter would be a few counts off and that worried me to no end.

Sullivan

This is a few counts out of how many now?

Weinreb

The electromechanical counter, I think, it was a 6 digit counter. It would be close to the end so say out of 1,000,000 but the number was pretty significant by the time you integrated that long. So I think I used a scheme that reversed the phase of the signal. This was a switched radiometer so I could make it so that drop counts would average out in the long run. The rest of the receiver was solid state. There were a few tubes in it in different places but it was mostly solid state. It was also an automatic balancing system. It had a noise diode whose temperature was automatically adjusted to balance the antenna temperature. And the front end was housed in a refrigerated box, a temperature controlled box that Ewen Knight bought for me. That is, Ewen made a contribution to the thesis which I think this box was $5,000 or so but it turned out to be a disaster of a box. The refrigeration system was very unreliable and when I finally went down to Green Bank, it failed. I replaced with fans and a porcelain control heating system but the front end was a tube, 417, or 416B I think it was. It was the lowest noise front end tube available at the time. System temperature, I believe, was around 1,000°. It was a 4 channel receiver. That is it would integrated when you were on comparison, another receiver would be integrating one signal.

Sullivan

I see. You had two front ends?

Weinreb

I had four front ends because it was dual polarization, dual polarization in all the time.

Sullivan

What was the purpose of the dual polarization?

Weinreb

It was all to gain integration time. That is, I knew I was going to need months of integration time.

Sullivan

[???]. It makes a big difference when you are getting into months.

Weinreb

Yes.

Sullivan

One other question I’m not clear on is your total bandwidth. Was that limited by the availability...

Weinreb

Correlator.

Sullivan

The correlator. Was that limited [???]?

Weinreb

Yes. The clock rate of inexpensive logic elements was 300 KHz, perhaps 500 KHz. I think the most I could run my machine at that time was a clock rate of 300 KHz which would analyze about 100 KHz band and I could analyze 4 100 KHz bands all summed together into the counters. I had a 4 channel receiver. The IF, all that, was transistorized in a nice little slide out drawer and I thought the equipment looked very nice. I paid a lot of attention to the appearance. Professor Bose at MIT helped me with it. That is Wiesner sort of didn’t have time to come into the lab and really do anything but Bose, who is the Bose of the Bose speaker systems, he was very nice about discussing problems.

Sullivan

Wiesner was actually your adviser?

Weinreb

Yes.

Sullivan

Well, how did the observations go?

Weinreb

Well, negatively I guess. That is, we got the equipment on the telescope. There was no signal. The fluctuation was going down the way it should; that is, I felt the equipment was working. I had a few ways of checking the equipment. One was an artificial deuterium line, a cavity that could produce an artificial line. I’m sorry it wasn’t a cavity. There were cavities in the front end as interference filters but I generated a noise spectrum. I could adjust how strong the line was in the spectrum. I generated it from low frequencies because it had to be a narrow line and, before I left MIT, the signal that I would get if the deuterium to hydrogen ratio was a terrestrial ratio I could detect.

Sullivan

After how long?

Weinreb

Probably a few days of integration. No, it must have been longer than that because I only got to half the terrestrial abundant. Well, I think that was probably at a 5 sigma or 3 sigma level. I don’t remember these numbers.

Sullivan

I can check these numbers.

Weinreb

Yeah, it’s in the paper. But nothing bad happened; that is, I didn’t get any false signals but I just was getting a nice straight line and the integration went on and on. I sort of broke it up into two week segments. There may have been a hint of a line in the first couple weeks. That just happened to be how the noise fluctuations worked but it just didn’t prove out.

Sullivan

You were looking at Cas A now primarily?

Weinreb

It was Cas A almost entirely. That is, I see now that was a mistake but I believed so strongly that I should look where there was strong hydrogen. It wasn’t going to vary. Wherever there was hydrogen there would be deuterium. I think we maybe spent a little bit of time looking at Sagittarius, maybe Cygnus. I was very impressed with Green Bank, the help I was given, the telescope operators. All the facilities were there: the telescope, the operators, base lab, base test equipment, very few astronomers. I think there were 100 employees to astronomer, [David S. ] Heeschen, Drake, and Menon but it was an ideal place for me to go. A new house to live in. Delightful place.

Sullivan

Well, you obviously decided to stay on pretty soon after that?

Weinreb

Yes, I guess. I was only going to go for three months. I then did a Zeeman experiment. I think I modified the equipment and got some help from Menon or Menon may have suggested it or helped me with it in some way.

Sullivan

This had been tried at Jodrell. I think by [Veuschuur?]?

Weinreb

Yeah. Davies.

Sullivan

Shuter and Davies, I think. But you could do it more sensibly?

Weinreb

Yes and I got another negative result which I think is correct that at the level I said. It was below a certain level.

Sullivan

Now, where did you look for that?

Weinreb

Let’s see. I believe that was also in Cas A. I might have looked somewhere else but I was pretty discouraged at that time with two negative experiments and people at MIT were dubious about the thesis I would say. I had a hard time with the thesis at MIT. That’s somewhat understandable. It was a new equipment technique and I had two negative results both of which there had been what we now know to be false positive results but they weren’t known at that time.

Sullivan

Had there been some positive Zeeman at that stage? I’d forgotten that.

Weinreb

I maybe wrong about that.

Sullivan

I’m not sure. I’ll check that out.

Weinreb

I thought that there were. So I had a hard time defending the thesis.

Sullivan

They weren’t maybe sure that your equipment was working to be blunt about it?

Weinreb

That’s right. I think I probably failed the first thesis exam or the first general exam or maybe both. I remember taking a lot of exams or taking some of them over again.

Sullivan

What did you have to do to satisfy them next time around?

Weinreb

Well, I don’t know. I may have gotten a letter from Drake saying that he thought it was a good experiment. I did have this simulated line result.

Sullivan

But you didn’t put that together as a result of the criticisms of your first [?]?

Weinreb

No, I’d done that initially. So they eventually graduated me.

Sullivan

That’s really amazing because I always considered that thesis one of the classics in radio astronomy.

Weinreb

Well, I realize now it was an excellent thesis, both the astronomy part of it, building the equipment, the correlation receivers have been used so much since then but I didn’t know at the time it was a good thesis and MIT didn’t either.

Sullivan

There wasn’t any one really challenging [?]?

Weinreb

There wasn’t anybody there. [Alan H.] Barrett came around that time. He probably helped. I think he may have been in on the second thesis exam. Lilley was on the thesis exam committee also. So I eventually passed.

Sullivan

It was ’62 that you passed?

Weinreb

Yes.

Sullivan

Let me just ask to summarize, what was it about the correlator? You say the basic techniques of the correlator had been put forth by Weisner and others.

Weinreb

Well, correlation techniques without clipping is the Wiener-Khinchin theorem, Norbert Wiener. The relationship between power spectra and autocorrelation function had been done by mathematicians many years ago- I don’t know, 1940s- so I knew that I could either build the processor that computed either the correlation function or the spectra.

Sullivan

Whichever turned out to be the most convenient [???]?

Weinreb

Yes and then I saw that it was easier to compute the correlation function digitally and then I saw that you didn’t many need bits. One bit was sufficient to do it.

Sullivan

But had correlators of a similar nature been built for other purposes, in physics?

Weinreb

For submarine sonar work. [Ferren?] and Hills had built a correlator. I believe it was an analog correlator.

Sullivan

And also there was the analog correlator by, what was his name who is now at Penticton. I think it was a development maybe simultaneous, maybe a little bit later than yours but rather independent. Ed Argyle did he not build an analog correlator about that time?

Weinreb

I don’t know.

Sullivan

So it was mainly in the application in astronomy, the clipping aspects, and making a workable system, of course, worrying about all the level problems and so forth and that was new aspect?

Weinreb

The digital processing was a new aspect.

Sullivan

Now, of course, one great advantage that one thinks of is the ability to change band widths so easily. Was that foremost at the time in any sense or was that sort of a by-product?

Weinreb

By-product. I think even in the original machine I could change the bandwidth over a 4 to 1 range.

Sullivan

That was sort of a motivation for you, to go for that sort of thing?

Weinreb

No.

Sullivan

Ok, so you had your degree, and then I’d like to talk about the OH experiment also. Was that the next thing or was there something else?

Weinreb

No, that was the next thing. That was primarily Barrett, Alan Barrett, pushed for that, suggested it.

Sullivan

He had looks at NRL [Naval Research Laboratory] many years before unsuccessfully.

Weinreb

Yes. I was sort of discouraged. I didn’t believe any lines other than the hydrogen line existed but I modified the equipment. It had to work at a wider band. I think I got it up to 250 KHz bandwidth and interfaced it with the computer at Millstone [Hill Observatory]. John Henry actually did the interfacing. I modified the correlator. This was a nice thing to have, the online computer. You know it was the first time we could actually really see the spectrum right away and I don’t remember what the front end was- whether I built that or something existed at Millstone. Anyway we put that together and I’m pretty sure it was the first night, we saw OH and I just couldn’t believe it. [L. M.] Lit Meeks was always an optimist. He was sure from the beginning that we were going to detect it and I was sure we weren’t. And a bump appeared in the noise and I thought it was noise but it was about the right place. It didn’t have quite the shape. It wasn’t a simple line but it was right there on the screen. It was marvelous.

Sullivan

This was Cas A again?

Weinreb

Yes.

Sullivan

But you knew the HI absorption was complicated in Cas A so perhaps that wasn’t that surprising that you would see more than one feature in OH.

Weinreb

Yes but it didn’t look the same. We looked again the next night and got the same bump. We still weren’t sure if it was interference or something. But it didn’t take long, you know, within a week we were pretty sure and I don’t remember whether we looked somewhere else, when we started looking somewhere else. I mean OH is much stronger in other spots that we didn’t look for. I think maybe the next step was finding polarization in OH. I don’t remember whether we did that.

Sullivan

Well now, of course, that had to wait for the discovery of the emission.

Weinreb

Yeah, ok.

Sullivan

Let’s see now how does the emission story go? That was Ellen Gundermann at Harvard and then at Berkeley. It was sort of a simultaneous thing to find the emission. As part of absorption studies, people began noticing this positive thing and then polarization came quickly. But were you ever involved in that, in the emission studies?

Weinreb

Yes, for several months there were few groups- we were one of them at Berkeley and I think the Swedes became involved also- there were sort of discoveries every few months: emissions, the line ratios weren’t coming out right, polarization, circular polarization. There was a lot of excitement.

Sullivan

Yeah, upper limits and also time variations were suspect.

Weinreb

That came later.

Sullivan

They’re still suspected. Those are the variations, I think, now most of them suspect because of pointing problems but, in fact, some of them were probably right. This is all in ’64 now, is it now?

Weinreb

Yes. I came to Green Bank as an employee in October of ’65 and OH was discovered in ’63 so most of this was in 1964. We started building a really nice correlator for the Lincoln Lab that was 100 channel, 10 megacycle clock rate correlator.

Sullivan

Which was used for a decade.

Weinreb

Yes, yes. Green Bank started building a correlator. Art Shalloway came to work from Green Bank, started doing that.

Sullivan

So do you think that that is a fair appraisal then that really up through the ‘50s that it was more of less building a different kind of antenna would be they way to make the discovery rather build a far better receiver?

Weinreb

Not just building a different antenna, just looking for something different.

Sullivan

But I would argue though that the antennas as soon as you build one that had a different capability either more sensitive- of course that’s virtually the same as having a more sensitive receiver- or a different spacing or a different way of looking at the sky, new features came that weren’t seen before. Now that you say the receivers...

Weinreb

That’s true to some extent but for example, the hydrogen line didn’t require building a different antenna. It was a simple form. It didn’t require any lower noise receiver than existed.

Sullivan

It did require some special techniques. I mean the frequency switch is not a straight forward thing and certainly novel.

Weinreb

Ok, Dicke switching came in World War II. It didn’t have any frequency switch to do hydrogen [?].

Sullivan

No, it didn’t have to.

Weinreb

No, I think it was more just knowing where to look. The technology was there. It had been there say since the 1940s. It was more knowing just what to look for.

Sullivan

The sky was so big that the cleverness came in knowing what to look for and where to look for it rather than...

Weinreb

Yeah.

Sullivan

But do you think that phase has sort of come to an end?

Weinreb

No, I think there is still a mixture of experiments of types that came about because of advances in equipment. Others just looking in the right place.

Sullivan

Ok, well thank you very much. That ends the interview with Sandy Weinreb on 31 July ’78.