Charles Supper Award Talk. August 1, 2012.
American Crystallographic Association Annual Meeting
The Westin Boston Waterfront
Transcript provided by Jenny Pickworth Glusker
Introduction by George Phillips, University of Wisconsin Madison, ACA President.
The Supper family, as you know, has had at least two generations of instrument makers — Charles, father, and Lee, the junior - who served the crystallographic community very well over the years with their engineering and manufacturing projects. I have the word "Supper" all over various bits and pieces in the laboratory and have had for a very long time, and they, the Supper family, saw fit to contribute to this award.
I am very pleased to introduce Ron Hamlin who is the recipient of this year's Charles Supper Instrumentation Award. Ron got his Ph.D. in physics at U. C. San Diego. He and Xuong recognized that modern electronics could be used for what film had been done at the time and devised very powerful multiwire detectors that essentially revolutionized the way crystallography was done, including the accuracy, the efficiency, and really took, particularly protein crystallography, to the next level. And about 1983 he, along with Xuong and Chris Nielsen, formed ADSC (Area Detector Systems Corporation) in order to commercialize this and disseminate the technology to the rest of the community. Those that were fortunate enough to have one of these multiwire detectors were the envy of all those who were waiting to get theirs, and this really did change fundamentally the way crystallography was done. A little later Ron recognized the values of CCDs as detectors and he and a number of people [were] working on this — Walter Phillips, Sol Gruner, Ed Westbrook — but it was Ron who put it all together in the next generation detector using CCDs. So, like Charles Supper and his son, Ron has provided the crystallographic community with a continuous stream of innovative tools, and I would like to ask him to come up now and to present him with the award for all the good work. [Applause].
"2-Dimensional X-ray detectors - what do we really want and how can we build it?"
Thank you. I am honored to be up here especially since I think it is probably rare that an instrument vendor is recognized like this, so it's really a special honor to receive it.
[Shows the audience his award certificate. Applause. Broad smile from him.]
So I think a number of people have seen me sitting in my booth at my PC doing this [he shows his fingers working at the PC] all week. And it's just that I don't give talks very often and so I'm hoping that I do OK, but bear with me.
[Slide 1] The whole subject I'll talk about has to do with X-ray diffraction.
[Slide 1: Header. 2-Dimensional detectors - What do we really want and how can we build it?.]
[Slide 2: X-ray diffraction. 1912 Max von Laue discovered X-ray diffraction by crystals when he and his assistants directed a beam of X rays at a crystal of copper sulfate and recorded the diffraction pattern on a piece of film. 1913 W. L. Bragg reported the first crystal structure, the structure of Sodium Chloride.]
And these were the two people who did the first two steps. [Slide 2] Max von Laue and co-workers shined X rays at a copper sulfate crystal and recorded spots on a piece of film. That was remarkable. But then W. L. Bragg found that this wasn't just a curiosity, but that it turned out to be quite a tool for solving structures of matter and so he figured out the structure of sodium chloride. That was the first reported structure and then that was the beginning of the use of this remarkable tool — X-ray diffraction.
[Slide 3] So this is the basic geometry. You have an X-ray beam, however you have created it (and there are many different ways). You have a sample and the samples that we will be talking about are crystalline samples, and you stop the main beam, but what scatters out of your sample are these diffracted beams and you record them somehow. First they were recorded on film, but then I spent my career figuring out ways to do something better than film. And so that's where all the electronic detectors came in that we developed.
[Slide 3: Basic Data Collection Geometry]
But this is the basic geometry and nothing really much has changed about it. [Slide 4] If you put a piece of film up there and rotate the crystal a little bit during your exposure you can get a rotation photograph, like that, and film was a pretty remarkable detector. And so what I will be talking about, by the way, I will be talking about what happened before I got involved in the instrumentation, because I think it needs to have some context starting very much at the beginning. And then I came into it about 1970 when detectors began to be electronic and when the area detectors were first developed. But there needs to be some discussion of what came before, and the thing that came before was film.
[Slide 4: An X-ray Diffraction Pattern recorded on film.]
If you built a special camera called a precession camera, and put a layer-line screen on it, there is a fairly complex motion of the crystal and the layer-line screen that would give you these beautiful photographs. [Slide 5] The advantage that these had is that you could index the spots right off the film. You didn't have to have a computer program to figure out what the spot indexes were and that was remarkable, that was the real advantage of the precession camera and the precession photographs that were made. I asked around. People used to estimate the intensities of these spots and that was the whole goal. If you got the intensities of the spots you could figure out what the structure of the sample was. People used to estimate these spots by having some kind of a little strip that in one way or another they exposed a piece of film for different amounts of time and then they would number them with the number of seconds or optical density or whatever. But they had some kind of a little strip that they used to compare their standard spots with the spots on the film and, remarkably, you could solve an X-ray structure with even eight sample spots. In fact, I was even told that you could solve it with three. You can have dim, medium, and bright, and that, if you kept track of your h, k, l indexes and recorded these spots, you could really actually get a structure. I think that is remarkable. [Slide 6] But that took time and I am told that people were even hired to sit and do this for hours and hours and days and days in order to get the intensities that were needed to solve the structure.
[Slide 5: Visual intensity estimation]
[Slide 6: But eyeball estimation of spot intensities took time and, at best, had limited accuracy. Was there a better way?]
[Slide 7] So the first step in doing something to make film more usable and more accurate was a film scanner. I found this in a book and I added [pointing to the slide] the red arrows to try to show where the beam was. The film was attached to this drum so that it could be rotated and then light was passed through it and the light beam was scanned along the film at various step sizes that you could choose, and then, so, the intensities of all the spots on the film could be recorded. I think the intensities that were recorded took eight bits so that is still better than one out of eight which is that only three bits, so that eight bits is even a step up. But you could scan film electronically, and that was a big step in getting the accuracy that was there on the film — getting it recorded. [Slide 8] This is a picture that Dieter Schneider got me. This is one of the really classic film-scanning machines, and he went out and cleaned the dust off this one at Brookhaven and took this photograph for me. I appreciate that he went out and did that. But this is, then, very much cleaned up. He said it was out by fertilizer sacks somewhere out in a storage house (I think they called it the boat house at Brookhaven). [Audience laughter.] Anyway he went out there and got it cleaned up so that he could get me this picture. But this is the first step in automating the data collection for X-ray diffraction experiments.
[Slide 7: Concept drawing on a film scanner.]
[Slide 8: Optronics P-1000 film scanner.]
[Slide 9: But data collection with film still required long x-ray exposures. And then developing film had to be done very carefully to get the accurate results that film scanners were capable of. An excellent paper on the careful, accurate use of film scanners is one by Walter Phillips and George Phillips. J. Appl. Cryst. (1985).]
[Slide 10: Electronic detectors. Early work was started in the late sixty's and early seventy's to automate the whole data collection process and use electronic detectors. This led to the development of commercial diffractometers.]
[Slide 11: Single crystal diffractometer.]
[Slide 9] There still had to be a lot of care to get film to get the accuracy that you really could out of the film. And George Phillips and Walter Phillips wrote a good paper later on about this and I think they probably wrote the definitive statement about how to develop the film and then how to scan the film to get the best accuracy you could from it. [Slide 10] But at the same time there was some work to just make electronic detectors and not have to depend on the film. [Slide 11] So that started and this was the first instrument that used an electronic detector, even though it was a small one, was the diffractometer and so I got this out of a book and put it up here, but in this case the sample was put in the X-ray beam and omega was the scanning axis. You slowly stepped through omega and then you measured the intensity of each reflection one at a time in this small detector.
[Slide 12] There's an example of a diffractometer that I remember seeing in Cleveland when I went to Picker in 1970 and talked to Tom Furness about it. He said this was pretty much the ultimate machine in the time in 1970. [Slide 13] The problem with the diffractometer was that you were only measuring one spot at a time and it took a long, long time to go in and to measure a high-resolution data set. It could take as long as two weeks to collect a data set. [Slide 14] The way that was done, as I mentioned, is you get a spot in the diffracting position, and you get it so that the diffracted beam falls in your little detector, and then you slowly scan omega and record the counts at each small angle increment and you can get a scan of a reflection profile. So I think this is, at least, this is where I first understood what a reflection profile is. That it was done one reflection at a time on diffractometers. Now we can do thousands of them (if we want to) at a time, but, in those days, one was what you got. And as we were just looking for the integrated intensity, under here, that was the number that was stored for each hkl that you measured. Many people were very careful to sample background, then bring the spot onto the Ewald sphere and go back and sample background. That was the most thorough, but it was also very slow. You had to do a lot of samples, so some people would just go sample the few at the top and just do peak sampling and do other methods for estimating the background. But still they only measured one at a time. [Slides 12 and 13 again]
[Slide 12: Picker diffractometer c. 1971]
[Slide 13: Diffractometer: automatic but measured only one hkl reflection at a time.]
[Slide 14: Omega scan of a single reflection]
Now 1970 was my entry point, so I just want to say here is what was available in 1970. [Slide 15] There were rotation and precession cameras. Film scanners had been developed and were working well, but the films still took some time to develop and, in fact, time to expose. Diffractometers were in use, but they were just measuring one spot at a time, but, at least, that's where things were when I got involved in it. [Slide 16] In 1970 I had just finished my exams in the Department of Physics (after you have to take two years of physics graduate courses), passed the exam, and then you can go look for a boss. So I went around campus and found that Xuong had an opening and he was working on things that seemed interesting to me, especially if and they had to do with electronics in some way, and I am interested in that.
[Slide 15: By 1970. Rotation and precession cameras were in use for obtaining diffraction photos on film. Film scanners were in use for optically scanning these films and extracting integrated intensities. Diffractometers were in use that were equipped with single point X-ray counters for automatically measuring one reflection at a time.]
[Slide 16: Also By spring of 1970 at University of California, San Diego. I had just finished the required 2 years of physics graduate courses and passed the physics departmental exams. In June of 1970 I showed up in Xuong's lab looking for a thesis project to start my Ph.D. work.]
[Slide 17] So here's what I got off the web. Xuong's computer crashed about a year ago, and all his valuable old photographs seemed lost, and so I had to get this off the web. But this is what I found. Xuong was educated in France and he got his Ph.D. at Berkeley. [Slide 18] Now that was a lucky thing because at Berkeley there was a lot of high-energy physics going on and Xuong knew the people, he knew some instrumentation and he knew how to program an IBM 1130 and write Fortran programs for it. All those things were important later on in our development. [Slide 19] This is just saying about what I said now. He had a background that turned out to be quite lucky for us to bring this high-energy physics idea into protein crystallography, so he was ready for that. I don't think he understood it until later.
[Slide 17: Xuong Nguyen-huu]
[Slide 18: Two critical things about Xuong's background. 1. He had worked for several years in High Energy Physics in the early 1960's at Berkeley just before he came to UCSD and so was familiar with some of the recent instrument development in High Energy Physics. 2. He knew how to write Fortran programs for the IBM 1130.]
[Slide 19: When I started work in Xuong's lab in 1970 he had already developed the screenless precession method for analyzing film data but he was asking the question How could we build a better (electronic) detector for Protein Crystallography?]
So when I started work there we started asking questions. Well, what could we do to improve on the diffractometer? We wanted an electronic detector, but one spot at a time was hard to like. [Slide 20] Again this is one of those times where some pieces of technology are available, but they are often not known to the people who need them. And so Xuong just happened to be bridging that gap. He had just worked in high-energy physics, and he thought it was a big science thing. He wasn't very interested in it. He wanted some kind of science where he could be more personally involved in development and so he had come to U. C. San Diego from Berkeley and decided to try to work on protein structures. [Slide 21] But two of the people he knew had done this work. Georges Charpak had developed wire chambers for high-energy physics and Victor Perez-Mendez at Berkeley campus (and Xuong knew him) had developed a way of getting position readouts by using delay lines. Delay lines were pressed against this printed circuit board and then this circuit board had many, many wires that went across this wire plane. They were fine wires, they were only 20 micron wires, real easy to break, but the central wire plane was held at 3,000 volts and two wire planes on the outside on each side were held at usually high ground and they were bigger, stronger wires. But Victor Perez-Mendez knew how to get position readout and Xuong knew that, and so Xuong started putting two and two together. [Slide 22]
[Slide 20: The right pieces at the right time. This was one of those lucky times in the development of technology when the important pieces of what we needed were all just becoming available. We just had to understand what they were and blend them together for a new application.]
[Slide 21: We drew on pioneering work from these two people Georges Charpak and Victor Perez-Mendez.]
[Slide 22: Georges Charpak had developed gas-filled wire chambers for use as position sensitive detectors in high energy physics (he later won the Nobel Prize for this) And Victor Perez-Mendez, a High-Energy physicist at Berkeley, had developed an accurate position readout method for these detectors based on wire wound delay lines.]
And so Xuong was also very good at persuading people to do things for him, which, again, is a personality strength. [Slide 23] He went to Victor Perez-Mendez and got him to agree to build us the essential pieces of the wire chamber, the wire planes, and the first delay lines that Victor Perez-Mendez had developed. And then we in San Diego agreed to make an enclosure for the wire planes and delay lines and build a mount for it to go on to our sealed-tube generator. [Slide 24] This is the geometry. This is an old slide. You can tell so when you try to scan this old slide; it did not come out very well, but this is the idea. A wire plane has three electrodes. There's the back wire plane which is held at zero volts. In this case we substituted a beryllium window for the front wire plane. But again it was held at zero volts, and the wires in the middle, the anode wires, were held at 3,000 volts. So, if an X ray came in, it was absorbed in the gas filling which, in our case, was xenon gas. Ions would be created in the xenon gas. They would migrate towards the high-voltage wire and there would be a pulse. There'd be an amplification right at the surface of the wire and there would be a sizable pulse of several millivolts that would be coupled in to the delay lines, one for x and one for y. In the case of the anode wire one there was just one wire that had the pulse and for the cathode there was a distribution of pulses that were induced by the event here in the anode wire. But there was at least a round pulse that was put in the delay line and then it traveled the distance over the readout that we could figure out the distance by measuring the time so we got a sharp pulse out of this one that started at time equals zero and then we measured two delays, one for x and one for y, and we ran two clocks, and we got an xy position that way. So each time an X ray hit, we could figure out its xy position, and we would add 1 to a histogramming memory. The histogramming memory turned out to be a scrap core memory that Xuong bought from RCA and it probably took me 6 months to get the thing to work. But finally, after I replaced a bunch of transistors in it and did some changes to it, I got the histogramming memory to work, and that was enough histogramming memory for two images from this so we could take images in one half of the histogramming memory and then change the high order bit and then take the next image in the other half while we were reading the first half out. So we had, and we didn't understand it at the time, we had shutterless data collection. It was just nanoseconds, so, in this case, it took one CAMAC cycle to change the high order bit from zero to one and so we had nearly shutterless data collection with a one-microsecond time between images but we didn't know that term at that time, we just had a detector that was electronic and could be gated.
[Slide 23: in 1972 Xuong convinced Victor Perez-Mendez to help us build a wire chamber detector for a protein crystallographic data collection using his delay line position readout system. The wire planes and delay lines were built at UC Berkeley by Victor Perez-Mendez. The gas tight enclosure and detector mount were built at UC San Diego in Xuong's lab.]
[Slide 24: Concept drawing of multiwire counter with delay line readout.]
[Slide 25] This is what a wire plane looks like. Some of these wire frames were pretty battle-tested. They would break wires and once in a while we would go in there and chop the wire out of the glue and add the glue and put a new wire in. So these wire frames were a real maintenance issue. But this is one that had been used for probably a year. This is what they looked like. The delay line was pressed against here [shows on slide] to do the position readout. [Slide 26]
This is what it looked like in cross section. We just built a big aluminum box to put the wire planes in. There's a delay line and there's another one down that way. The box is filled with xenon and the X rays came in through the beryllium window. [Slide 27] We got help from our friends at U.C.S.D. Joe Kraut's lab had an old G.E. quarter circle goniostat that they loaned us. It turned out to be just perfect, because it didn't shadow much in the diffraction pattern. Just a quarter circle and, when you rotated it, it didn't cast a big shadow on the side where we had our detector. So it turned out to be really lucky event, and it was free. They just had it sitting in their storage, so we took pieces of it and mounted it like this. [Slide 28] So here is, this is what we call the Mark I system, and this was a big event for all of us because, in this case, you didn't have just a single point counter, you had an area out here where you could measure quite a few spots all at the same time. Similarly you could still do profile scans, but you could do profile scans of many spots.
[Slide 25: Wire frame used in the Mark I detector.]
[Slide 26: Cross section of Mark I.]
[Slide 27: An old GE quarter circle diffractometer was donated by Joe Kraut's Lab at UC San Diego and we mounted it on the table top of a Picker sealed tube X-ray generator.]
[Slide 28: The Mark I Diffractometer at UC San Diego in 1975.]
So here's the Picker sealed-tube generator tower. There's the G. E. quarter circle goniostat that was given to us, and here's the big Mark I wire chamber. It's important to remember that their spatial resolution wasn't so good. This was 128 bits. 128 wires that way and then we could resolve in time and we could resolve 256 that way. So, our pixel for 1 mm. by 2 mm., which was ridiculous by today's standards, but really a thrill back then. [Slide 29] So, instead of being able to measure just one spot, like a diffractometer did, now we can do that. [Slide 30] And that, in its day, was just remarkable. We could measure 50 to 100 spots all at the same time. In a similar way we could just measure all these numbers. We had an electronic detector that could count X rays, but we could measure many of them at the same time. [Slide 31] I think the next image is with the detector put straight on. But that, to us, was a big thrill in its time. It looks really primitive, the pixels are big, but it was really something to be able to measure that many spots. This is a crystal of subtilisin. I think its space group is P21 and the unique axis is vertical so that we could get up-down symmetry. And we usually adjusted the crystal position till we got pretty good up-down symmetry and then collected data that way (with the up-down symmetry adjusted).
[Slide 29: Diffractometer: automatic but one spot at a time.]
[Slide 30: The Mark I system covered a big piece of the diffraction pattern.]
[Slide 31: Diffraction pattern from Mark I.]
[Slide 32] So, that was where I got my Ph.D. At the end of that we collected some data from dihydrofolate reductase that was being grown in Joe Kraut's lab by Dave Matthews. He was generous enough to loan us some crystals he had done uranium soaked. So the major point in my Ph.D. thesis write-up was a peak in the difference Patterson, a uranium peak in the difference Patterson. Now we were really thrilled with it - that was a real crystallographic result from this new kind of detector. You'd think that a uranium peak ought to be pretty easy to see, because it's got so many electrons, but, at least, we could see it. It was good and clear and the dihydrofolate reductase didn't last long in the X-ray beam. So it was actually a good match to a much more efficient data collection machine that we had built. And so Dave Matthews also, I think, was in the right place at the right time. When a machine like this could collect data from a single one of his crystals and that would not have been possible from a diffractometer because the crystal didn't last long enough. This was another just lucky break for all of us.
[Slide 32: Ph.D. Physics, University of California, San Diego, 1975. Title of my Dissertation. The Multiwire Proportional Counter Used as a Position Sensitive X-ray Detector for Protein Crystallography. One of the important crystallographic results reported in my thesis was a good strong uranium peak in a difference Patterson map generated from data collected on the Mark I multiwire Diffractometer from Dihydrofolate Reductase (crystals grown by Dave Matthews in Joe Kraut's lab at UCSD.]
[Slide 33] Here is the overview of what I've been doing for forty years. Instead of having film in here, we worked on electronic X-ray detectors. But I want to keep in mind that the big picture is that you have something that makes X-ray beams, and there is a lot of technology there, you have something that helps you grow samples, and there is a tremendous amount of technology there. We built X-ray detectors. But then there are a lot of people who write sophisticated software for analyzing raw images, for generating electron-density maps and then generating structures. So, this whole long process needs to be done with the goal that is over here. So we were just a piece of the process and I think I have to keep that in mind. I was more like just an instrument builder and I didn't really understand very well all the rest of these pieces. But this is my box, this is the box that I've been in for forty years, right there. [Big smile.]
[Slide 33: X-ray Based Structure Solution. The Big Picture.]
[Slide 34] So we could get 50 spots but, there were a lot more spots out there than 50 spots. Well, we thought, what can we do to improve on that? OK, let's at least build two detectors - make a two-detector system — and we should be able to get more spots. [Slide 35] And, we learned how to build the wire planes. At first the Mark I wire planes were built at Berkeley in Victor Perez-Mendez' lab, but we built our own wire winding machines. [Slide 36] This is Don Sullivan, this is me with the longer hair I had in those days and we really learned how to build wire planes. And we did find a company that wound us a few of the delay lines. That's a whole other subject too. [Slide 37] But we ended up building a two-detector system. So this was a big improvement — we got more solid angle that way. [Slide 38] Now we could see that many spots - and that was pretty good. I want to go back here and point out one more thing. Notice that these detectors are on big heavy stands. This was all we did in those days. [Slide 37] There was a big screw thread on here. You could adjust the detector heights, and down here was a protractor we had to draw on the floor, so that we would drag these stands around and set them at what we called "two-theta angles" and that way we could position the detectors in various places in the diffraction pattern, depending on what we wanted to do. [Slide 38] We usually put one detector so that we could see the beam-stop shadow so that was a low-resolution detector. Then we put the other detector up at higher resolution. That was a typical setup for this system. [Slide 39] So the two-detector system was really a step up and also (I didn't mention that) we switched [to] the rotating anode and that, although these things are a lot of maintenance, it was a lot brighter and so we got more beam intensity (we still had a monochromator) and two detectors.
[Slide 34: But we wanted to improve the detector system some more. An obvious next step was to use a pair of detectors instead of just one to intercept more of the diffraction pattern.]
[Slide 35: We learned to build all the parts of the multiwire detectors in Xuong's lab.]
[Slide 36: We were careful to do it right.]
[Slide 37: The two-detector Mark II system started operation in Xuong's lab in about 1982.]
[Slide 38: More solid angle, more reflections.]
[Slide 39: With two detectors and the increased beam intensity of the rotating anode X-ray generator, heavily replicated data sets could be collected in 10-18 hours, often from only one crystal!]
[Slide 40] So that system was pretty respectable in its time, and it became a Research Resource funded by the NIH. Protein crystallographers could apply for time. I'm not quite sure who made the decision. [The official protocol was that Research Resources advertised for beam time and applications were reviewed and awarded for perceived merit.] I think it was just Xuong. I think if you knew Xuong and you made the right comments to him, you got time. But I think they were involved in that, but somehow (I'm not sure how competitive it was), but Xuong knew who was doing work that he was interested in and he would ask people to come in, usually, I think people came for about a week. They were often graduate students. He would set up a group of people who would come. [looking out at the audience] Judy Kelly was there, Marv Hackert was there. They'd come for about a week and collect lots of data and go home, I think usually pretty happy with the results.
[Slide 40: This instrument in Xuong's lab became an NIH research resource about 1983. Protein data collection time was awarded to groups who submitted proposals to use the machine.]
[Slide 41] I stayed around from the time I got my Ph.D. in '75 until '84, just a little bit after we started the company. And we continually improved the multiwire system. We worked on all kinds of electronics, especially electronic improvements to make the images smoother. And, of course, my other job was maintaining the Elliott GX-6 rotating-anode and that was hard, that machine was a kind of primitive machine. It's a bright machine but it is primitive. It broke a lot and we kept having to resurface the anode and change the seals and so forth and so forth. But that's another story. That was part of my job. So I stayed around for those nine years in Xuong's lab as a research scientist.
[Slide 41: From 1975 to 1894 I stayed on in a Research Scientist Position in Xuong's lab at UC San Diego as we maintained and improved the Mark II multiwire diffractometer system.]
[Slide 42] But after people came to this Research Resource they really kept saying to us; "You guys should make commercial ones of these, or somebody should. There should be commercial availability. Why do we have to just wait our turn to come to San Diego (although it's a nice place in San Diego). Why do we have to wait our turn and just come once in a while? Why couldn't we have one of these things at home?"
[Slide 42: But ... There was a clear demand for a commercial version of these multiwire counter detectors.]
[Slide 43] So, without much knowledge, or any knowledge of how to start a business or run a business, why Xuong (my boss) and Chris Nielsen, who's here — he's the software expert in the Xuong lab — and I decided to just take a chance and start a company. In fact we bought a book that said "How to start your own California Corporation." You filled in the blanks and you could tear these pages out and send them in to California. We did that and then we started our corporation out of the book of tear-out sheets and then we showed this to the lawyer that we went to months later and he said "No, no, no, you'd better let us rewrite this." So, for thousands of dollars more they did what they thought was a more professional job. But the one that started our company was the tear-out pages out of that book. Anyway, we started the business. The first location was in my apartment. [Slide 44] So there was the workshop. In fact we still have this drill press back in the corner of my shop and we still use this toolbox. This was in 1984. This was in one 100 square foot bedroom in my apartment near U. C. San Diego.
[Slide 43: In 1983 Xuong, Chris Nielsen and I started a company called Area Detector Systems Corporation. First business location: my two-bedroom apartment near San Diego.]
[Slide 44: First workshop in my apartment in 1984.
[Slide 45] There is the first multiwire system that we built, that is serial number 1, and ended up in — it was bought by David Eisenberg. He had been one of the data collectors at San Diego, he liked our data, he wanted to help us start a company, and so he gave me a purchase order for $50,000 which I thought was a tremendous amount of money to build a system and deliver it to his lab. One of the mistakes I made in that is just what happens when you aren't accustomed to running a business — I forgot about sales tax, and so the state of California wanted their sales tax and so I had to pay it. I should have put it in his purchase order but I forgot. So I paid $3,000 of sales tax — just by total surprise. That's how little business experience I had. I didn't really have the money. This was a little company. I had borrowed $20,000 from my Mom and Dad to start the company and the $50,000 in income was going to help us get our business going, and it was not pleasant to have to pay someone $3,000 as sales tax. But that's how it turned out.
[Slide 45: Parts of first ADSC multiwire counter system in my living room in 1984.]
[Slide 46: This detector, S/N 001, installed to David Eisenberg's Lab at UCLA in September 1984.]
[Slide 46] There's that detector serial number 1 [S/N 001] installed in David Eisenberg's lab. Again we had it on these stands. These were just commercially available camera stands and then we made this adapter bracket to put our detector on. You still had to roll these camera stands around and set up an approximate two-theta angle and then the computer had to refine the position of the detector, but it was usable and was an area, and so now it was an area detector, an area diffractometer, so that was the first one in Eisenberg's lab at UCLA.
[Slide 47] Then, in 1985, because it was just ridiculous to run such a business in my apartment, we moved to a little [?], what they called an "incubator space." It was 800 square feet in a little business park where a lot of other small businesses were starting and so this was our first location. We still have that sign somewhere in our shop. We had to have the sign. So we had this little carved wood sign hanging.
[Slide 47: 1985. ADSC moved out of my apartment and into an 800 square foot space in a business park in San Diego.]
[Slide 48] And we started building better equipment. [Slide 49] There's a wire-winding machine that we designed and built ourselves. We had learned some things about the ones that had been used in Xuong's lab and we designed a better one with four bars instead of two, out here, to put the wires in. The threads in here gave you wire spacing. [Slides 50 then 49] And there's the same paddle, the wire planes were put on here, and the wire was wound around it, and then, [Slide 50] after the wire was wound on, we would slide the wire plane under it and then solder each individual wire to the plane circuit board there. [Slide 51] And then we also bought a coil-winding machine, but this machine was huge and it was made from winding big transformers. So we had to make some special parts here to wind this tiny #38 wire that was used in the delay lines, and this is something that my wife Haruko, who is right here, she wound all the delay lines for all 83 of our detectors, so she would sit here, sometimes I think she could probably wind one in about a day. She had to keep stopping and pushing the wire tight, so the wire winding had to be perfectly tight, 3,000 turns on one of these delay lines. But this is the machine that we had and we had this special stuff made here and a magnifier so that she could see what she was doing during the wire winding. So we customized this giant coil-winding machine especially for making delay lines.
[Slide 48: We set up the equipment we needed to build multiwire counters. We built our own wire-winding machine to string the fine wires on the wire planes. We bought and modified a coil winding machine so my wife Haruko could wind the delay lines that were used for the X, Y position readout. We designed and built the readout electronics.]
[Slide 49: The wire plane-winding machine we built.]
[Slide 50: A wire plane on the paddle of our wire-winding machine.]
[Slide 51: Coil-winding machine custom modified for winding the very small, #38 gauge wire required to make the delay lines.]
[Slide 52] This is what the delay lines look like. You can see, here's the number "# 38 wire" and she would wind 3,000 turns of that on a form like that, on a printed circuit pattern that was important to the delay line work. She would wind those on there and, even then the system delay here was about 2 microseconds and we would measure time and figure out the position along the delay line. That's how we got our coordinate. But these were not really very linear still. We needed ± 1 nanosecond in 2 microseconds so we developed a way to put extra little capacitive strips on the delay lines to linearize them to ± 1 nanosecond. We never published this but it's important to get smooth images, especially from the anode wire plane direction where each pixel had to be exactly one anode wire spacing wide. But with this capacitive patching that we devised we ended up getting good linear behavior, so our 2 microseconds was ± 1 nanosecond over that length. And it had to be, or the picture wouldn't be smooth from the detector.
[Slide 52: Delay lines for the position readoff.]
[Slide 53] In 1987 David Eisenberg bought a second detector from us. One of these is serial number 1 (I can't tell you which one). Serial number 1 is the early picture. And then they bought serial number 16. We made a two-detector system, put it on a little bit fancier camera stands that could have two-theta arms, and then Duilio Cascio and I at UCLA drew this protractor on the floor so that we could approximately set the detectors out at some two-theta angles.
[Slide 53: A double-protractor system UCLA late 1987 S/N 001 and S/N.016.
Over the years, then, I drew probably a dozen or more of these protractors in other systems that we installed. I still have my plumb bob and some of my tooling that we used to lay these protractors out. But that was an important part of our two-detector system. But protractors were not easy to draw and lugging the detectors around by hand, pushing them back and forth to set the detector angles, really wasn't very convenient and so we thought couldn't we just motorize all of this? [Slide 54] And so this is a classic picture. This has really put Marv Hackert's lab in our map forever. This is in Marv Hackert's lab. This is serial numbers 13 and 14 of the multiwires and he actually built a cinder-block wall. He took some of his raised floor out in that area and built cinder block walls to support this. This was built by Dave Rognlie's machinist at Blake Industries. We went to them and said, "Hey, we need a motorized two-theta table," and they built quite a wonderful thing so you could drive the two-theta around. The distance was still manually adjustable, but you could at least drive the two-theta accurately. You see you get a pretty good match if you put a crystal up there and get a diffraction pattern. Your spots are pretty much where they should be within a millimeter or so. This was quite an advantage in its day.
[Slide 54: Motor driven two-theta table. U Texas 1988 in Marvin Hackert's lab.]
[Slide 55] We built many, many of these detectors. In fact, at one time, we had a two-year backlog of purchase orders. These are detectors that are probably in the serial numbers in the middle 70's. We built 83 of them and this was in 1990. These were detectors with serial numbers around 75, somewhere. We got so that we could build these things pretty well. We had a real factory going. Gas chromatograph, very important to study the purity of xenon gas. In fact this was a real breakthrough. We needed that gas chromatograph. [Slide 56] So, this is a list, as much as we can reconstruct, of all the systems that were delivered. I know David Rose is here, I've seen him, he had this system there. It looks like serial numbers 71 and 72.
[Slide 58] At the same time we were selling detectors there was another guy who had a similar idea, and I owe him credit. He was our arch competitor. Either we sold our detector or he sold a lab his detector. The detector he developed was Xentronics detector (as it was called at first). It was smaller, it was more compact, and would fit on a tabletop. And, in some ways, it was quite attractive to a lot of labs who wanted a smaller compact system, and so I give him a lot of credit for being a good competitor, and a good guy. [Slide 59] His detector had a spherical beryllium window in front so that you could put xenon and a quench gas mixture at a pressure of (I'm guessing) four atmospheres, but at high pressure you needed the curved window to support against the pressure. So this detector had really good stopping power and if the sample was scattering diffracted beams, they would all then be completely absorbed and then the ions would turn and go into a regular xy wire chamber. [Slide 60] This detector did well and it had a good solid angle coverage. I'm guessing because I never used one of these, and George has. Maybe the intercepted circle was bigger and depended on the crystal quality. It's a good piece of solid angle that you could get with these. The only criticism that people had was that you had to have the detector in pretty close and the spots hadn't spread out as much and so the background hadn't spread out as much and so maybe your peak-to-background ratios weren't quite as good with a small detector, but for the convenience and a pretty good solid angle coverage why, these guys were stiff competitors to us.
[Slide 55: Six freshly-built ADSC multiwire detectors in our shop in 1990.]
[Slide 56: From 1964 to 1992 ADSC built and installed 83 multiwire detectors.]
[Slide 58: Ron Burns - a worthy competitor — developed the Xentronics detector.]
[Slide 59: Concept drawing of Xentronics detector.]
[Slide 60: Typical solid. Xentronics detector.]
[Slide 61] But what we needed was a detector that could even measure more spots and, unfortunately, here's a case where we just didn't do it right. We didn't realize that a new technology was coming that was really going to shut us down. So we made a business decision here and it's one I hope we'll never make again because we didn't do R & D ahead of the detector we were making. We thought this was the best detector in the world almost and why does anyone need anything better, why should we work on R & D. Well, were we wrong! [Slide 62] Image plate detectors were being worked on and we ignored them. We should not have.
[Slide 61: But what was needed was a detector with even better spatial resolution and even bigger solid angle coverage.]
[Slide 62: Such a detector had been under intense development in Japan and then in Germany. Image plate detectors began to be commercially available.]
[Slide 63] So this was one of the first image plate detectors that I knew about. And some of this is my personal story so I don't plan to say much about Rigaku image plate detectors, but this was the first one that I had anything to do with. So this detector pretty much shut us down. Our purchase orders stopped, some of the backlog of our purchase orders, some of those people cancelled, and they went off and bought these. [Slide 64] But the reason, and this is what it looks like inside; there was a laser system up here, and a photomultiplier tube, and, after an exposure, this laser system would come down and read the image off the scanning system. [Slide 65] I'm sorry, but this is a crude picture. But the idea was that a laser was directed at the round image plate and it was spun fast and then a track (like the needle, like a phonograph record groove) was scanned. Light was given off in proportion to the X rays that had hit that part of the image plate during the exposure. Light was given off and measured, and so a good image could be gotten. [Slide 66] But the point here is now look at the solid angle coverage and that's what did us in. It still took two minutes to read one of these things out. It was a bit of a step backward because it was an integrating detector, not a counting detector, but the solid angle carried the day. It could measure a huge solid angle. And it was simple. You didn't have to plan multiple sweeps to get all your spots measured, you could just collect data at one setting and just rotate your crystal and get all the spots. So that was compelling and it ended up just becoming the dominant system. [Slide 67] So, because our purchase orders stopped, and because we hadn't done any R & D to make our own image plates, we were approached by MAR and they offered to let us be the U.S. and Canadian distributor of the MAR image plate scanners. And we agreed. At least we could keep our company going. So that was our damage control reaction. OK, we will sell Jules Hendrix's image plate scanner because we didn't do our R & D and we got caught by surprise. [Slide 68] So in these years, and these were some of the most difficult years for me personally and our company, we sold somebody else's detectors. But they were popular and I think we delivered about thirty of them. We didn't make these detectors. All I did, I hired somebody who went out and did installations, we just imported the detector, calibrated it in our shop, and delivered it. We didn't make it and we didn't probably understand it as well as we should have. But it was at least a way to save our company.
[Slide 63: One of the first image Plate Systems c. 1991.]
[Slide 64: MAR 180 with cover removed.]
[Slide 65: Concept drawing for image-plate detector.]
[Slide 66: Image Plate Detectors. Brute force solid angle coverage.]
[Slide 67: In 1991 ADSC contracted to distribute the MAR Research image plate scanner systems in the US and Canada. From 1992 until 1996 ADSC installed about 30 MAR Research image plate systems.]
[Slide 68: 1992-1996. Image plate detectors quickly came to dominate in home labs and especially at synchrotron beam lines because not only did they have very large solid angle coverage but they were integrating "film-like" detectors so they were not count rate limited like diffractometers and multiwire counters and were particularly well suited to the higher intensity diffraction at synchrotrons.]
[Slide 69: But the one-to-three minute readout time of image plate detectors was still too slow to get the best use of a synchrotron beam line. What was really needed at synchrotrons was a detector with much faster readout time.]
[Slide 69] The one-to-three-minute readout time was still a problem and, at synchrotrons, although these detectors were used a lot at synchrotrons because they didn't have any count-rate limits, they were integrating detectors, but you really wished that you had a detector that would read out faster than that because the synchrotron was just sitting there, waiting to start the next exposure, while you took minutes to read the image out. [Slide 70] So, as George mentioned, these guys had been doing early CCD work - Sol Gruner, Walter Phillips and Ed Westbrook — and so what we did, knowing that we were real beginners at CCDs, we went to Sol and Walter and asked if they would be our consultants. And it turns out that I think that's the thing the company should keep doing, is look and see where the pioneering work is and go talk to those people and try to be the first commercial company that talks to those people and then try to commercialize what they have, and never stop doing that. Just keep watching the horizon to see what's coming next. And go talk to those people as soon as you understand which ones they are.
[Slide 70: In 1994 encouraged by the pioneering work of Sol Gruner, Walter Phillips and Ed Westbrook, ADSC started work on commercial CCD-based X-ray detectors.]
[Slide 71] So here's the basic idea of a CCD detector. Now there's a phosphor. The X rays, they strike a phosphor, they give off light, it's green light. But there's quite an energy loss here going with 12 keV. You only get this many visible light photons, but you lose a lot of energy here, and then only some of those photons actually get into the fiber optic taper, and, if 450 go in, a lot less go out. There's a huge loss of light that gets scattered inside the fiber optic taper, so only that many come out, and then the CCD sees, maybe, only that many of these, but it's enough. If you get eight electrons per X ray you can see individual X-ray photons. The CCDs are low enough noise, the good scientific grade ones are low enough noise that you can see one X ray, if you only get 8 electrons, out of the original huge amount of energy that came in the X ray. So they're good enough. [Slide 72] This is just to show us what a fiber optic taper is. [Slide 71] It is similar to a lens in some ways, but you have to press the object that you want to transmit the image of. You have to press it against one end and then the thing that's going to see the image also has to be pressed or glued against the other end. [Slide 72] So they have these bizarre properties and, I think, they're neat things just to move around out on the table. And they work in either direction. You can use them to magnify something or demagnify something, but that's the way they look. They're just made of millions, literally millions, of small fibers that are put in an organized way so that the image stays pretty well organized when it comes through. [Slide 73] This is the kind of taper that we actually use in our detectors. We had them. They were made round and we had them cut square at both ends. This was a demagnification ratio of more than three. So we could put 105 mm out here, and put the phosphor against it and glue the CCDs back here. So that's how you could match a fairly small sensor to a fairly big area where your phosphor is.
[Slide 71: Basic principle of operation.]
[Slide 72: Fiberoptic tapers.]
[Slide 73: Single fiberoptic taper cut square for use in an array.]
[Slide 74] We started making these detectors. We called them Quantum 1s. This is the first Quantum 1, and there's a taper, and there's the readout electronics. We went to a company that made these electronics called Princeton Instruments in Trenton, New Jersey, and so our first CCD detectors used their preamp electronics and their controllers and we just did the packaging to make little Quantum 1 detectors. [Slide 75] We sold about 16 of these. Many of them were sold to small-molecule applications through Rigaku. We do better when we sell directly to customers than when we sell to people who do packaging because we lose track of who the customers are and so much of our business is personal relationship between we who build stuff and the customer who uses it, because then we get to hear how it works and we get to hear suggestions. So selling it to somebody else who has this as their system just isn't what we like to do. But we did some and that helped us run our company at the time. [Slide 76] The good thing about the CCDs in those days was 9 seconds. It didn't take two or three minutes to read these things out. You could read them out in nine seconds, and that was pretty impressive, we thought. It's not, by today's standards, but it was pretty good. [Slide 77] So we made a bunch of these Quantum 1s. As I was saying, many of them are packaged up in small-molecule systems and sold through Rigaku, many of them in Japan.
[Slide 74: Quantum 1 cover removed.]
[Slide 75: An ADSC Quantum 1 detector.]
[Slide 76: An ADSC Quantum 1 detector.]
[Slide 77: An ADSC Quantum 1 detector.]
[Slide 78] But protein data collection synchrotrons need a bigger area. [Slide 79] So we thought OK, well, why don't we just make an array detector and put four of these things together. Now that turned out to be 188 mm on a side and that was pretty good. [Slide 80] That's the detector, that was our first array detector. We sold the first one of those in 1996 to ALS (Advanced Light Source), Berkeley to Thomas Earnest. In fact, during the development of the CCDs, we just barely had enough cash to get through. I put my personal savings in the company at that time to keep the doors open. We got this order from Thomas Earnest. He gave us a cash advance on two of them and that rescued us. We were as close to crashing as we ever had been. But we learned a lesson — watch your cash, and then you can get your first orders. Try to get cash advances, but you have to get advances from people who trust you, and I think we had earned enough trust that Thomas Earnest decided to give us a chance. And then we delivered both detectors. [Slide 81] Here's what else we did. Once we could build Quantum 4s we decided to promote them, and we went all over in the world. This one was at Brookhaven X12B and this was a big event in our company too, because here was a $400,000 sales price detector, and we just loaned it for at least a week, maybe longer than a week, to collect data there. But, compared to the image plates we used on the beamline at that time, it was just amazing, 9 seconds readout, and that was one of the images that we got and said "Look at the spots in that thing," and you can get that every nine seconds, so you could expose in nine seconds and you could get that. [Slide 82] This was a crystal brought from David Eisenberg's lab by Duilio Cascio and he was just amazed. So this really caused him a buzz when he got going. These CCD detectors are really something at a synchrotron. [Slide 83] We sold quite a few Quantum 4 detectors. These are the installations. Most of them were to synchrotrons. A couple of them were to [cough] #419 went to David Eisenberg's lab.
[Slide 78: But for protein data collection at synchrotrons we really needed a CCD detector with bigger area.]
[Slide 79: Quantum 4 CCD Array Detector.]
[Slide 80: Quantum 4 photo.]
[Slide 81: Quantum 4 demonstration at X12B, June 1997.]
[Slide 82: Quantum 4 demonstration at X12B, June 1997. Glutamine synthetase from David Eisenberg's lab at UCLA.]
[Slide 83: 31 Quantum 4 detectors delivered 1996-2001.]
[Slide 84] But still nine seconds. Well, couldn't you do that better, because the synchrotrons would even want a faster readout detector. [Slide 85] So, OK, let's go find a better CCD. We changed from a CCD that we could read out in 9 seconds (It could only be read out from a single corner) and went to a different CCD that could be read out from all four corners. It also had more pixels and it was bigger, so we could also make our tapers bigger. So we built a detector that could be read out in one second. In fact the binned data could be read out in something like a third of a second (two by two binned data). So we went a little bit bigger but mostly we went from nine seconds to one second. We went quite a lot faster. [Slide 86] So the two-by-two arrays of that detector were called the Quantum 210. We sold quite a few of those. [Slide 87] But here again you keep thinking couldn't I just do more, more. Well, more. The next "more" was couldn't I go bigger. [Slide 88] Now that I've got faster couldn't I go bigger? [Slide 89] So we thought — well, we've made two-by-two arrays. Let's make three-by-three arrays. Now that's a serious size — that's big enough for protein crystallography.
[Slide 84: But for synchrotron data collection we wanted even faster image readout.]
[Slide 85: Quantum array detectors.]
[Slide 86: Quantum 210.]
[Slide 87: 18 Quantum 210 detectors delivered 2001 - 2008.]
[Slide 88: Now that we had faster image readout we wanted still bigger detector area.]
[Slide 89: The Quantum 315 uses 9 instead of 4 of exactly the same modules as used in the Quantum 210.]
[Slide 90] So there's the next detector we made called the Quantum 315. There are a lot of those around. [Slide 91] There's one of the first ones that was delivered to Stanford in 2001. They built their own positioning device that could raise and lower the detector and tilt it so that it could essentially simulate a two-theta angle. [Slide 92] And, here's one of the images from that. Now that is really a lot of spots. That's something like 15,000 spots and, in fact, the crystal here is so good, it has such a big unit cell, it almost cries out for a much, much bigger detector. This detector in full resolution has 6,000 by 6,000 pixels and, in fact, the screen resolution here doesn't really show how many spots there are. In fact, Jamie Cate at Berkeley was good enough to give us this. He said all that we should say about it right now is that it's called bacterial ribosome. He's still working on this. He doesn't want to give out more information there. And this image has only one pixel out of every six in both dimensions. That's all that he wanted to give us right now. It's good enough to show a lot of spots but he didn't want to give us something that we could index. [Audience laughter] Then, to his credit, this is a competitive business but he was generous enough to give me this about a week ago just to show the number of spots that you can get on a CCD detector. There is something here that's significant. See that the low order spots are saturated and CCD detectors are only so good and they do saturate when you put too much intensity on them? But, in this case, I think he must have gone through in a separate pass and measured the low-order stuff. But he was trying to get all of this out at a higher resolution. So I thank him for that. This is from James Holton's beamline at ALS. I think Jamie collects most of his data on James' beamline.
[Slide 90: CAD drawing of Quantum 315.]
[Slide 91: Quantum 315 at SSRL.]
[Slide 92: Quantum 315 diffraction image. Bacterial ribosome. Jamie Cate and Jonas Noeske. James Holton's beamline 8-3-1 at the ALS.]
[Slide 93: Diffraction pattern from a Quantum 315.]
[Slide 93] So this is a comparison, this is kind of a historical look-back. If you superimpose what we could do with our multiwire counters with what I roughly estimate an image plate might have done, image plates might have come out bigger. But this is a comparison compared to one spot. We could do maybe one spot from a diffractometer. We could maybe do 100 spots in our Mark II multiwire system. Image plates could do 100 to 1000 spots. But this is 15,000 spots. So that's what changes have happened over the years that I've been working on this. [Slide 94] We delivered and had delivered 32 of these Quantum 315s, all to synchrotrons. And we did a whole lot of traveling all over the world, had some good repeat customers. [Slide 95] But now [looks at watch] [Slide 96] We are coming to pixel array detectors, and I'm going to have to speed up.
[Slide 94: We delivered 32 Quantum 315 detectors 2001-2012.]
[Slide 95: But especially for the new brighter synchrotrons being built we need detectors that have 1. even faster readout and 2. much larger dynamic range than CCD detectors have.]
[Slide 96: Pixel array detectors are now becoming the dominant X-ray detector technology at synchrotrons.]
[Slide 97] Pixel array detectors are the next thing. They detect X rays in a layer of silicon. The charge (which for 12 keV) can be quite a lot of charge. It goes down through individual bumps in the processing circuit down here. [Slide 98] This is another picture of what they look like, but the problem with pixel array detectors is they have to have bond wires that go back to the readout electronics, so there are gaps in pixel array detectors because these bond wires have to come in to the front of the readout electronics. [Slide 99] But they have very high efficiency, they have low noise, [Slide 100] they do not saturate easily because they have giant dynamic range. [Slide 101] Usually the processing circuit has a counter or else it integrates charge. [Slide 102] This is a big one that everybody knows about. You can see the gaps, these are because of the bond wires.
[Slide 97: Basic structure of pixel array detectors.]
[Slide 98: Photo of a pixel array detector.]
[Slide 99: Pixel array detectors: Important features. Direct detection of X-rays: High efficiency with low noise contribution. Each pixel has detection logic and counters and/or integrators built in. The ASIC signal processing layer can be built efficiently in 2cm by 2cm size and larger areas can be built using tiled modules.
[Slide 100: ... And more. Very fast readout time (~1 millisecond). High frame rates (~1kHz or higher) are possible. Shutter-less data collection for crystallography! Extremely high dynamic range (ability to record very strong and very weak data simultaneously). Sub pixel point spread. Weak signals are not affected by a neighboring pixel's strong signal. Flexible pixel design allows for detectors to be customized for specific scientific applications.]
[Slide 101: But pixel array detectors do have gaps... So all larger multi-module PAD (pixel array detectors) detectors have small dead spaces (gaps) between the modules.]
[Slide 102: PILATUS at Swiss Light Source. Pilatus pixel array detector showing gaps.]
[Slide 103] There are three kinds of pixel array detectors. The simplest one just integrates charge in a pixel. The next level up in complexity is one that counts X rays. And then there's one that we call charge ramp counting and that's one that we developed.
[Slide 103: ASIC: Application specific integrated circuit ASIC's are custom designed silicon chips, There are three basic kinds of signal processing ASIC's in pixel array detectors: charge integrating, photon counting, and charge ramp counting.]
[Slide 104] This is the circuit for the integrating pixel, where charge just comes in, X rays hit this diode layer and a charge comes down through the bump bond it goes into a capacitor and, depending on how many X rays hit the pixel, you get more charge at the end of the exposure. You just digitize the voltage on the capacitor. The problem with these is that they can only store about 10,000 X rays in a pixel. [Slide 105] They are simple, but the capacitor fills up pretty fast. But, for fourth generation machines, by the way, they're the only thing you could use, because you can't count fast enough in a 50-femtosecond pulse. You have to use purely analogue pixels for a fourth generation machine. [Slide 106] This is what the Pilatus does. They get a signal, it's amplified, compared to a pulse-height threshold. If the pulse is bigger than a threshold you get a count in your counter. Counters are typically 20 bits or more. [Slide 107] The advantage is, again, they are simple, they run at room temperature, they have some energy at that resolution because of the pulse-height thresholding. But, at high count rate they do have some count-rate limitations. And, for each different energy you have to readjust the pulse-height thresholds in all the pixels.
[Slide 104: Charge Integrating Pixel.]
[Slide 105: Advantages. Simplicity of design. No pulse height thresholds to adjust, and purely analog so usable at 4th generation pulsed X-ray source with extremely high instantaneous flux. Disadvantages. Capacitor fills up with charge after about 104 12 keV X-rays so limited dynamic range.]
[Slide 106: The Photon Counting Pixel.]
[Slide 107: X-ray photon (pulse) counting. Advantages. Simplicity of design, room temperature operation, and some energy resolution. Disadvantages. Count rate limitations, and pulse height thresholds need to be reset for each different X-ray energy.]
[Slide 108, Slide 109] So this is the technology that we developed, based on ideas from Sol Gruner. You could use a little capacitor and you could let the charge build up on your little capacitor up to a point, up to a threshold, then you could remove a precise amount of charge from your capacitor and just do this again and again and again and count how many ramps of charge on this little capacitor you got, and then you can have a big counter over here that is ramps instead of X rays. If a ramp is a charge of about 100 X rays and if you had a counter over here that, say, does 20 bits, that's a million. You can count 100 million X rays without saturating the pixel. That's way more than 10,000, if you just had a single capacitor, so this charge ramp system is really powerful for handling much higher exposures without saturation. [Slide 110] And, in fact, because you can count ramps at a megahertz and because they are evenly spaced, you have no coincidence loss counting ramps. You can handle 100 million X rays per pixel per second with no coincidence loss. Since it is not a counting detector, there is no coincidence loss. [Slide 111] At low count rates, however, there may be a slight advantage counting X rays. And then you have the pure counting statistics that come with individual X rays and that might be better.
[Slide 108: There is a better solution to higher count rates than a simple counting pixel: The charge ramp counting pixel.]
[Slide 109: The Charge Ramp Counting Pixel.]
[Slide 110: The main advantage of the charge ramp counting pixel is that it can accurately measure up to 100 million X-rays/pixel/second with no co-incidence counting loss.]
[Slide 111: But at low count rates — less than 100,000 X-rays/pixel/sec where co-incidence loss is not a problem... counting individual X rays will probably give superior measurement accuracy and can allow some energy resolution.]
[Slide 112] So the question is, can you have pixels that can do either one? [Slide 113] And so that's what we are developing right now. This is a more complex pixel. We have silicon now and we have tested.
[Slide 112: Can we have it all? Very high effective count rate in the bright parts of the diffraction pattern and The ultimate accuracy of counting X-rays in the weaker parts of the diffraction pattern.]
[Slide 113: We are developing an ASIC with pixels which can be configured into either of the two modes. We call these pixels "dual mode" pixels. Can we have it all?]
[Slide 114] So these pixels can either be put in pulse counting mode like the Dectris Pilatus, by closing that switch and leaving this one open, [Slide 115] or they can be put in ramp counting mode and then you can count charge ramps. You have a much higher dynamic range and you can't saturate. So, in fact, we've gone up to 22 bits there, so that's four million. Each ramp is 100 X rays. Why, that's 400 million X rays before you have saturated that pixel in that exposure. And then, at the end of the exposure, there's a ramp always in progress and so you can digitize that residual ramp to 10 bits. And at that point, since you've got 10 bits and this is only about 100 X rays in here you have about 10 A to D levels per X ray and, depending on noise conditions, you can probably see an individual signal from individual X rays. You get 32 bits of data out of this mode, and in the silicon we have it's just a bit in the configuration register in each pixel to tell it whether it should be in pulse counting or ramp counting mode. [Slide 116] So that's what we call a dual mode.
[Slide 114: The dual mode pixel in X-ray pulse counting mode.]
[Slide 115: The dual mode pixel in charge ramp counting mode.]
[Slide 116: We have just completed testing of the 16 by 128 pixel version of the dual mode ASIC and are ready to submit a foundry run for the full sized 128 by 128 pixel production version]
[Slide 117] We have a contract with Wayne Hendrickson who is good at finding us some development money to build him a 2K by 2K detector with these dual mode pixels. The first thing we will deliver (each one of these ASICs is 128 by 128). So this is a 512 by 512 proof of principle detector that needs to be delivered to him at his old beam line at NSLS late this year. We are estimating November, but we will have a dual mode proof of principle detector delivered to him by then. [Slide 118] This is the one that Wayne wants at his beamline, but I don't think that the NSLS II beamline will be ready for this so we'll have to deliver it to NSLS X4A, so he can at least test it. These are the specs: 150 micron pixels; the area, but it does include gaps, is 328 by 308 mm, so it's about the size of the Quantum 315. We will be able to get at least 100 Hertz frame rates. There is more about why we can do that. In fact, in certain clocking modes we could probably go drastically faster than that if we only want fewer than 22 bits per pixel. But, I have a talk later this morning and I'll give you more of the technical details of that.
[Slide 117: The first proof of principle detector we will build with this new ASIC will be the 512 by 512 pixel HF-262k to be delivered to NSLS beamline 4A about November of this year.]
[Slide 118: The scheduling of the NYSBC contract requires delivery of the
2K by 2K HF-4M detector to X4A at the NSLS by late summer of 2013. 150 micron pixels, 328 mm by 308 mm active area, 3mm horizontal gaps, 100 Hz frame rate.]
[Slide 119] In protein crystallography we would tend to, if this is the 2K by 2K detector, we would tend to program the pixels in the middle (where the diffraction pattern would have bright low-order reflections). We would put those pixels in the charge ramp counting mode because they couldn't be saturated, and then for the somewhat better statistics in the rest of the detector in the weaker part of the diffraction pattern, we would put those pixels into photon counting mode, but when we read this out we read it out with all with the same clocking and it's up to the software to determine what those counts are. Are they X-ray counts, or are they ramp counts? Then we would either pay attention to the ten-bit A to D conversion number or not, depending on whether we read them here or out there. So the readout would not have the care and only the software later on would figure out what the data means. You could program the detector in about 10s of milliseconds to get any pattern of photon counting or ramp counting mode.
[Slide 119: The dual mode PAD [pixel array detector] detector. 4 by 16 array of 128 x 512 pixel modules.]
[Slide 120] So here's my last slide, and I'm close to time. I've just seen all this happen in 40 years since I went to work for Xuong in 1970. A lot of things have changed. Synchrotrons have gotten better and better, detectors have gotten better and so, with the pixel array technology that we are developing now, after a protein data set can be collected in just a few seconds, instead of weeks as it was on the diffractometer. So that's the conclusion. But I've seen that happen in 40 years.
[Slide 120: What a change in the last 40 years! In 1970 a protein data set could take weeks to collect using film or a diffractometer with a single point counter. With the pixel array technology being developed now, accurate protein data sets can be collected in just a few seconds!]
[Slide 121] I'd like to acknowledge some people that helped me back in the early multiwire days. In the Xuong lab it was Chris Nielsen who did all the software. Carl Cork, Andy Howard, Dan Anderson, they were graduate students in Xuong's lab. Don Sullivan was the master technician. He taught me a lot about how to build things. Wayne Vernon was another high-energy physicist in the Physics Department. He and Xuong knew a lot about instrumentation. In the Chemistry Department Joe Kraut, a protein crystallographer, gave us some of the old equipment, Dave Matthews was his post-doc, Karl Voltz was his graduate student, Jeff Bolin was his graduate student, Tom Poulos was his post-doc.
[Slide 121: Special acknowledgment to the many people who contributed to the early work on the multiwire diffractometer systems at U. C. San Diego. Xuong lab: Chris Nielsen, Carl Cork, Andy Howard, Dan Anderson, Don Sullivan; Physics Dept: Wayne Vernon; Chemistry Dept: Joe Kraut, Dave Matthews, Karl Voltz, Jeff Bolin, Tom Poulos.]
These people all helped us when we needed crystals or advice on how to collect data. Thank you. [Applause.]
See the 2016 Update article on Ron Hamlin's 2012 Supper Award talk