1+1=3 or How I Learned to Stop Worrying and Love Holistic Circuits - A. Hajimiri - 1/29/2014

the following program is brought to you by Caltech you so I'm good evening I'm more agree lice provost for research and also my faculty of engineering at Cal Tech and I would like to welcome all of you to our second Watson lecture of this quarter and before I introduce our speaker tonight I would like to bring your attention to our next Watson lecture which will be given by professor Dennis coachman and it's titled everyone who starts small how metals learn to behave so that would be on February 12 of course ok getting back to our lecture tonight as you know our speaker tonight is Professor Ali Hodja Murray he received his BS degree from sheriff University of Technology and his master of PhD degrees from Stanford University he has been one of the he has been on the faculty of Californians for technology since 1999 1998 where he is the trommel's Myers professor of Electrical Engineering and professor of medical engineering and director of microelectronics lab his research interest includes integral integrated circuits and systems for applications in sensors biomedical devices photonics and communication systems and prior to become a professor at Caltech he worked at philips semiconductors and Sun Microsystems and Bell Labs and in 2004 he was selected to the world T r35 as top innovators the list of that basically every year is announced by MIT he is also a fellow of I Triple E Institute of Electrical and Electronics Engineers and has served as distinguished lecturer for I Triple E he has 65 granted US patents and many more pending actually applications and he has co-authored about 150 reviewed paper in top academic journals as well as author the book in quartered six chapters the different books and he has won many awards for best papers and basically in different fields of electronics Ally actually co-founded axiom Micro Devices whose fully integrated CMOS PA has shipped close to 200 million units perhaps you have one of those in one in your smartphone's right now well on a personal note note I'm very proud that I have been part of the team that we lured Ally instead of going to Stanford come to Cal Tech and Ally is actually one of my heroes at Cal Tech because he really represents a tourist spirit of our Faculty's entrepreneurial nature not only in starting new businesses but also in their approach to research to give you an example and Alice spent a year recently after becoming a full professor or usually we go take a big rest he took a course actually one-year course or training in biology in order to equip himself with necessary knowledge to allow himself to analyze new and challenging medical problems he actually took courses he sat side by side by our undergraduate students and asked the instructors actually to correct and give him the same harsh the time they give to every other students and he was very keen on that one and the time that he invested believe me is spending a year just focusing a wire firmly your research in some new fields is not something that every Caltech faculty will do so in that respect he it was a big investment for it for him but the result has been amazing because as you will hear tonight among everything else he has done he's come up with ideas but for new diagnostic devices that one day may actually detect cancer cells and other diseases so very easily by the patient themselves with this introduction of like all of you to welcome Ali to give his lecture tonight thank you well good evening everyone and thank you very much for that very kind introduction Marie it's a pleasure to be here tonight and being able to talk to you about some of the work we've done and share some of the results with you tonight's the title of my talk is a strange title in the sense that it's not mathematically accurate as probably everyone knows and the spirit behind that as well as some of the examples will show that it's really a way of thinking about the whole being more than its parts and really that really goes back to the concept of holistic circuits the way we'll just and I will give you several examples of that but before I do that I'd like to start with a little bit of a historical review of how we are where we are in terms of integrated circuits so I'm going to spend the first 10 minutes 10-15 minutes going over the history and the history of integrated circuits actually is intertwined with the history of computation and storage and really as you can see basically if you start from the very ancient times the history of ancient computation starts way back more than a hundred thousand a hundred thousand years BC with the computational tools the most basic ones we have which are our fingers of course these fine specimen are a little bit younger than that but in general that's there those are the tools the most basic tools that use now going forward basically this the next tool that makes things possible and made an advance was really the ability to store and keep track of the actual computation and storing and results for example the one that you see here on the in the middle top right the top middle I'm sorry would be is it basically a multiplication table made by Sumerians and this is pretty old actually if you think about this more than 4,700 years old abacus another mechanical device which is very useful then we went to storage on various forms of papers basically papyrus here is another example of the methods of storage and this isn't is a remarkable papyrus in the sense that it has multiple mathematical theorems and problems that were used by the students to study this one is much much more recent but I decided to include it because it was developed in the pen it's a mechanism for it's more of a record-keeping device it's an Incas people and what it is is really they use the knots here to keep track of the numbers so the numbers here basically are stored in the location and the number of knots that they have on each one of these strings and what this does basically is another device for storing mathematical information moving on to more recent times what we really have is that we are there's a tree observe a transition from a mechanical to electrical and this is an important transition we're starting off with things like slide rule this is a very useful device for making multiplications and divisions using logarithms and this is a very useful thing in that sense it was it was used up until the mid 40s and 50s and even you know I was curious as I had the one actually just wanted to learn how it worked when I was younger and then the other thing this is actually a Babbage difference machine Charles Babbage came up with the idea around 1860 1850 1818 50 1840 timeframe and what it was if he never got to build the whole thing it was a machine that basically calculated the value of polynomials by completely mechanical means using completely mechanical so he had to full design but he had he had a disagreement with his machinist then it at the end basically did they parts that he was demanding from him and the precision that he was demanding was not something that basically he could give him and they had dispute so basically it wasn't built up until more recent times based on his complete design but it's purely mechanical then this is another device basically you have a tabulating machine that was used in the census of 1890 with punch cards and going forward this is probably the first all-digital computer of our time which was used with electric which was made with electrical relays these electrical relays basically are mechanical devices that have electrical input and electrical output so we are transitioning from mechanical to electrical this was built around 1941 by Konrad Zuse a Z 3 relay computer and this ENIAC is the first one that was built the full computer built with vacuum tubes so it gets rid of the large mechanical moving of the devices so looking at this this is actually a one of the those interesting transitions that we see and this is an interesting curve this is an interesting calculation I think it's based on the work of reycarts will cares well and basically what it shows is a plot of computations per second how many calculations you can perform per second for for a thousand dollars for a grand how many computations can you get and this is basically a plot of that versus time and you can see that you have to spend a lot of money for certain number of computations back then and once you had the relays if the current became faster and the vacuum tubes made it faster and you can even see a change in the slope it became the transition even became faster and this is the so-called Moore's law of course of going for so he calls it the fourth Moore's fifth paradigm and these are the devices involved basically going from mechanical to electro mechanical to electrical but using vacuum and these is basically the way things were going till a more recent time until a wonderful device was invented now I'm sure everyone has heard the name of transistors probably almost everyone has heard it and what that is really is that it's a solid-state device it's a device that allows you to go from transition from a vacuum based system or system that's not completely solid to a solid-state device this device in my opinion was invented by Julius Lilienfeld in nineteen twenty five thirty in a sequence of patents that he actually filed and this is an example of the one that was filed in 1928 and was granted in 1933 and this describes a MOSFET it's really a MOSFET which is similar device to what we use today and the patent is pretty clearly written it's very well written and it describes the device really the way it is but he never there was he never published an actual article showing this device to work so there's a lot of question about whether or not he actually made one irrespective of that the first transistor that was really demonstrated publicly is this device and I took this picture when I was working at Bell Labs myself and this was made by these three gentlemen for which they won the Nobel Prize of 1956 and I would like to think that this is I would consider it one of the top ten in of all time but of course I'm biased so so in that sense it's an important device respect so moving on the next step was obviously to try to include more and more integrating have more and more integration so the idea was to now if you can make one of these devices what if you can make multiple ones what are the things that you need to do to put them all together now this was the basic idea is not that far-fetched I mean it's something that you would think that yes of course you want to put more and more of them on the same substrate on the same piece of semiconductor but the question is how there are three fundamental problems the way I would think about it was one was how to make them in parallel which was done by a process which is akin to printing press it's a parallel process a batch process the other thing is how to keep them separated isolated from each other because if they start interfering with each other then what you really get is not a collection of independent devices they're all attached to each other and they wouldn't work really the way you want them and the third problem is how to connect them to each other metallization and these problems were solved by various people some of them were named and some of them were given more credit than the others but in general this mode this was basically one of the examples of early integrated circuits that are designed and built so this was basically you this is an or gate that was used for the Apollo state spacecraft so basically that for an Apollo project this was used and the entire size of this market to give you a sense in 1962 was about four million dollars and so just gives you a sense of how think how rapidly things have grown now of course once you make things put things on the same substrate you make what make it on and make them on one substrate then the next natural question next question is that what's the next step what would you do if you could put more than one transistor on one substrate if you have to then the next step would be to put more right and that's basically Moore's law to put more and more on the chip so the question is how far can we go and what are the things we need to do really to put more devices to cram more devices on it's the same substrate you have to make them smaller that's one thing you need to do and this is the one of the reasons for the scaling proper scaling concept the scaling is a process by which we are making the transistor smaller and smaller and the main driver for this although what some people would like to think that was really integrating more and more the main driver for this is something else that we'll talk about in a second which was the speed read but now these are some examples of some of the commercially available devices today so for some of the commercially devices today these are pretty small if you want to think about it the channel width of a typical transistor today which is like you can see the cross section here this is about 70 atoms Y and this is a commercially available device it's not really something researching that much and you know it more interestingly the gate width of this thing is like three atoms wide and atoms are small I mean I know that everyone knows that atoms are small but atoms are really small believe me so that was one of the things but the other interesting thing is that if you want to make something faster if you want to make something move faster what do you do with it think about it if you wanted to make something faster more nimble it has to be lighter and smaller the speed of these transistors really is determined to the first order by the time it takes an electron to get from one end to the other end if you want to make it faster if you make it shorter then if it goes faster I mean really first order zeroth order way of thinking about it and that was I think the main driver for scaling which also had this byproduct of making it denser and being able to pack more and more stuff on the same chip so now where are we now so this is an interesting plot so let me tell you what this plot tells you this the x-axis is that obviously the year and the y-axis is the number of transistors on a single chip and these are real this is real data so this is around the time I was born which is close to where the first microprocessor was introduced this is very close within a year of my birth and this mic refers microprocessor a wonderful device Intel 4004 had about 220 300 transistors on it and we've gone up and there's no stopping it this part of Moore's law is not stopping and I believe actually this is going to speed up it's going to become a little bit steeper because the other parts are slowing down this is the number of transistors we can put on a single chip this is a single device now to give you an idea this is the four thousand and four and these are the relative sizes of these things are to the scale so and this one is that in Intel Xeon Phi 63 64 cores so this is the relative size it's you can see that the size hasn't increased that much which means that the transistors have become much much smaller much denser now the interesting thing is that another kind of a point is that Xbox one for example you can go and buy one for 500 bucks and the main chip on that device has 5 billion with a B transistors on it and that's not the largest ship we are getting very close to the 10 billion and we are to keep going we will keep going so this is a very very different kind of ballgame keep in mind the number of transistors on these chips are comparable to the number of people on this planet on one single chip these are mega ultra Cities they're building and for this chip to work every single one of those transistors have to work not a single one can fail just keep that in mind now the other thing that and this is based by the way these things are not that big as big as they appear on the screen so this is the relative size of that chip compared to a penny so it's not that big it's basically less than an inch on the side that's three 2/3 of an inch I'm sorry about that 3/4 of an inch on the side so moving on the other thing that happens when we make things smaller is that they become faster now this is a similar plot in terms of x-axis but on the y-axis that the cutoff frequency the maximum frequency at which you have gained out of a transistor some sort of an amplification I will talk about this briefly a little later now this is also a logarithmic plot you can see when you go 1 1 step up you basically multiply by a factor of 10 so the transistors have gone from this point to that point so basically their speed has increased by more than a factor of thousand in terms of how much how fast they can operate and where you can get gate now this is the part that's kind of our kind of slowing down you can see it and that's because of the physics of the device and this is the one that I believe will force the number part to go up because if you can't make them faster use more of them essentially that's the philosophy but now the other interesting thing is that we when you do this when we make these transistors faster by making them smaller it didn't come exactly for free the price we paid is that we went from bulky muscular devices like this which are discrete devices of course to very very delicate yet agile devices so we made them very small very fast but they are puny so the question is that if we have a situation where we have an unlimited number of transistors at their disposal no limit but they are very very weak individually what do you do with that what's we have to change our mindset and this is not such an unusual mindset in terms of like people have thought about this mindset before this is from of course The Wonderful Wizard of Oz and this is a painting from this is a drawing from the original edition and what you see here is that you really have two different paradigms the army of mice versus the big lion we've gone from the lion to an army of mice and the way you use an army of mice is very different from the way you use a liar if you had a lion at your disposal you would use it very differently and there are things that an army of mice can do that the lion can never they can get places that the lion can never get they can do different things at the same time and that's what we mean by a holistic perspective of design in the sense that we really have to use this large number in a very different way utilize a different aspect of the physics of the device as well as the nature of the device itself so how do we do that well I'm going to give you several examples of this thing and the examples we are going to start with an example which is using the context of waves electromagnetic waves so we'll do a little bit have to take a little detour of way on waves and we'll come back to our actual images and what we actually did so let's go and talk about waves of course waves we are familiar with the waves we've seen the water waves you know the ripples on the on the ponds and everything and all of those things are very interesting to us they have interesting behavior and if you think about it waves are our primary means of communication and sensing for mid and long range the two primary modes of long and mid-range communication for humans is vision and hearing which relies on electro magnetics and acoustic waves so you're relying on waves every day I mean you are hearing and seeing me because of the waves so the waves have an interesting set of properties right so one of the interesting properties of waves that's quite interesting Square quite remarkable is a coherence so let's talk about coherence what do you mean by coherence well to talk about the concept of coherence let me run an experiment with your help I will need everyone's participation in this so let's let's see what how it works so why don't we say the rows to the right of the auditorium so let's do one thing first as a group so let's everyone think about your favorite one digit number and say it out flied out loud on the count of three okay two three okay what does the number no coherence right so now let's try something else let everyone on the count of three say my favorite number five okay one two three that's coherence alright so what happens when you have two waves that are acting basically are coherent they going together up and down together like the red one and the blue one if the add they give you something larger but the interesting thing about this that has the twice the amplitude is that it has four times the energy so you get four times the energy and if they were going exactly out of phase they would give you almost nothing so does it mean that one plus one is four terms of energy or does it mean one plus one is zero in this case well it depends if you're adding coherent Li or anti coherently or incoherently this would be anti gray well this is of course not one plus one four I mean this is really if you want to write it correctly mathematically you have to write it this way because their power and energy is proportional to the amplitude squared but other than that and these are mathematically correct so there is nothing really there's no need to go back and revisit or basic principles of mathematics but the concept is here is that once you get these two to work coherently you get more than the energy of each one so what what happens does it violate the conservation of energy do you get more energy out than you put in and we don't have to worry about the warm energy problems and that solves everything no unfortunately not so what is happening what's happening is that you get more energy at some places but you get less at some other places so now imagine you have two sources to generate ripples on a pond or any kind of wave and they're going like that so they are going together and what you see here is a simulation of what happens so these are the two sources and you can see that in this direction you're generating let's take going straight up you're generating waves but and you so are you in this direction in that direction but if you look there these two channels where you don't transmit much energy so you get more energy in wandering in these directions and less energy in other directions so the total energy that you get is constant but you concentrate them in a certain area more than others now why is that important because now imagine that I could get the coherence and basically I could get everyone to act that way with a larger number of sources so what if I have more sources that are acting coherently going together helping each other so this is with 8 now if you look at the simulation you see that most of energy is going we're going upward and you don't send much energy in other directions so essentially when you are doing this you're forming and the more sources you put here the narrower this beam becomes so I can actually send more and more energy to a narrower sliver of space and I can create this wave that moves only in this direction going vertically right now now what is even more interesting is that now imagine that if when you were saying my favorite number I wanted to hear you all of you guys say this a the five at the same time but if I'm standing on this side of the stage now I will hear the voices of the people on the other side in the far end a little bit later than the people on this side right what if I got them to start a little bit earlier so all of your voices would arrive at my location at the same time then you have created beam not only you have created the beam but you have also focused it in my direction and that's what this simulation shows now you can see the one on the right is going a little bit earlier than the one on the left and what it does it has created a beam not only it created a beam but also has steered it to the left so this is the basic basis for what we call the phased array a system of multiple sources that individually can be small but collectively can add up to a large signal power and this is something that can be used to generate a lot of power and not only generate it and focus it in one direction but also steer it in different directions so keeping you that in mind this is another way you can also do that in on the receiver you don't have to just beat the transmitter now down the receiver side if you think about it let's say I want I'm interested in the signals arriving from a certain direction right arriving in this direction now if I adjust delays correctly I can add them all in phase and coherent but now if I actually go and try to look at something else and something let's say some another way for that arrives at it from a different angle I can actually suppress it by adding them out of phase and this way what happens is that essentially you get its suppression of the things you don't want and amplification of the things you do want so you can basically focus which way you're listening now the good news is that we all own of these or at least a pair of them forming a phased array this is probably the way and this is the way that you know if you are in a cocktail party and you're kind of like pretending to politely listening to somebody and you're really listening to the more juicy conversation going on on the other side you are you really using your phased array it's a two element one but it still works and actually if you calculate the dimensions for the acoustic waves and the center of the audible range it actually the numbers work out so it's not that it's completely crazy so and you know we were so excited about this line and this is this concept by the way it's not a new concept this concept has been around for a long time but we were kind of like one of the demonstrators so this is with when we went out with some of the grad students and they enjoyed it because they came out of their caves and we went and had a field day so basically this is this is the wavefront if you look at it here we created a wavefront the way we did it is that we gave each one at each person a number and he said one two three four five six seven eight and if at the time of when your number was called you basically created a wave and you can see that you created a wave in this case that went to the left so yeah we were having fun with that but that that's that's like a way to demonstrate something that's already known so we did this with what kind of waves with water waves for example or we talked about acoustic waves but electromagnetic waves are also waves and electromagnetic waves cover a broad range of the things that we interact with and deal with right I mean if you think about this is the broad spectrum of the way electromagnetic wave you started long wavelength low frequencies like this these are radio frequencies that you use for let's say AM and FM radio and then you go to microwave frequencies that you have in your microwave and your cellular phones are operating at those frequencies then you have this range in the middle of terror Hertz that we'll talk a little bit more about infrared visible light is electromagnetic wave ultraviolet you go to x-ray gamma ray and all of the range so it's they're all essentially the same thing at different frequencies with different wavelengths so they're all waves at different frequencies so and this is an interesting part of the spectrum which is pretty hard to generate it has very difficult to generate things in this range this is a range of frequencies that's basically called a terahertz gap between point three terahertz to three terahertz and there are many applications for it because this allows you to do fine resolution systems you can use it for for example human machine interface you can use it for touchless gaming nowadays you have we have systems like that but they are not very accurate this way you can actually detect the movements of the island and things of that sort you can detect the eye movements you can use it for very high-speed communication you can use it for six years security imaging you have such systems at airports today but they're bulky and very expensive the question is that can be make them very compact and cheap and then you can use it obviously for medical imaging and applications now the interesting thing about the terahertz radiation is that it being at a frequency where basically there is non ionizing photon energy and this is a more of technical detail it they are safer they cannot induce chemical change directly and what they do they can actually only warm things up now if you warm things up if you warm it enough then you can induce chemical change through cooking them but other than that they don't intrinsically create chemical change unlike things like x-ray so but the challenge so this would be very useful there are many applications for this but this involves making things that operate above the cutoff frequency of the transistor you remember the frequency plot that you were talking about early on and let's talk about that cutoff frequency a little bit more the idea of what is a cutoff frequency of a transistor so so let's this is the symbol for the transistor by the way if you're not an electrical engineer there is no really need to know that but of a MOSFET and now if you have an input at a low frequency a wave at a low frequency coming in this transistor amplifies it so it makes it larger right it makes it bigger at frequencies below the cutoff frequency now at the cutoff frequency essentially what you get out is comparable to what you put in and what this is really is that this is the definition of cutoff frequency in a band which frequency when you go to very high frequencies what you get out is smaller than what you put in so one could argue that a piece of wire would be more useful than this transistor because it at least doesn't attenuated it's a little bit more subtle than that and the subtlety really is that yes you don't get gain but there is still power and this really goes back to non-linearity in a more technical detail of this thing but what it is really is that the waveforms don't really look like the way I showed them like pure sinusoidal and they look like this and that means that there's energy at higher frequency so the question is that can we come up with a way to harness this energy take it out and radiate it out so we came up with this idea actually my former student professor kasha Sengupta he's a professor at Princeton now and we actually focused on this place mostly akashic worked on this and one of the things that we did was that essentially we created this structure which we call the distributed active radiator and what this really is is a method of using the finite speed of light to create the right phases in the system by using a large number of transistors and devices to create a structure that radiates electromagnetically very efficiently and it uses multiple transistors multiple small transistors to combine power and generate it now once you make these one of these you remember the concept of phase directs we had multiple sources right if you have multiple sources then you can play all sorts of games by controlling the relative phases of these things so now you can make an array of these things and they will each radiate certain amount and if you combine them coherently you can get all the benefits of being forming and being steering and we made a chip like that based on that so this is basically a 4 by 4 array of these things these are the individual radiators the da RS and this is the entire system this is about 2.7 millimeters to give you a sense it's about 1/8 of an inch on the side this is small now that's it's that chip in the middle it's that thing so this is all the cast and crew supporting there pre-madonna here and basically what it is it's just radiates directly out and it forms this beam and it's gonna steer it so does it work well yeah this shows these are measurement results and what these plots show is that you can form a beam this blue curve and you can steer it to the right and left up and down so you can go left and right up and down which basically is the basic things that you need to have for being steering and then we said okay well if you have this if you want to have a false you want to make a camera a detector for this so if you make multiple detectors if you make an array of detectors on it on another chip this is another chip you can make a terahertz camera each one of these you can think of it as a pixel well now this is not really a very high megapixel camera this is like a 16 pixel camera at this point but it's scalable you can add more and more pixels and get it to a point where you can get higher resolution right and then we said okay can we image stuff with this so we started imaging things so these are real images of some of these things that we actually did so this is an example of a key inside an envelope we know that goes through soft matter so you can actually see this is a drill bit inside a plastic casing so you can see through objects by looking at this this is a a pass for Subway I believe from Atlanta we were in Atlanta for a conference and then picked it up and this one is the antenna you can see the RFID chip then tell everything inside it this is a bag of sugar its sugar and no sugar and then here are some other medical kind of some more biological matter leaves dry and you know fresh and then you have chicken tissue different kinds of tissue will have different kind of absorption fat muscle and if you have more basically you have the diff you can differentiate between bones cartilage muscle and fat using this kind of errors imaging over very thin layers of matter now and then you said okay well if you put all of them together can you make the entire system with a silicon thing the point of making this in silicon chip using a CMOS technology is that these things in volume will cost very little because of that wonderful scaling that we have because of that huge number of transistors that we can make these things in volumes would cost less than a buck so you can think about putting them on your phone or other things or their applications that and that's a completely different question of like what are the other things you can do but there are many many many applications for something like that and these are some actual images for the entire system and this was something that we did as a demonstration of security imaging so this we had this is the app Jeffy's imaged and this is the image we got and this is what's inside that object alright so here's the question so the next going forward so that was an example so we did make these phased arrays in very high frequencies we've done it actually the RF frequencies microwave frequencies millimeter wave frequencies but then the other frontier is that next one up next frequency up is light visible or near visible light the question is that can we do this with light because if we could it would be bending the light right can you bend the light so let's see so the idea here is that if you want to make a system a phased array again but in optical frequencies using up silicon photonics chips and the way it would work is that you feed it with a laser or multiple lasers and then what it does it forms a beam with that laser and now what you can do is that we can electronically control by controlling the delay so you can electronically control where this beam is pointing imagine my laser pointer but now imagine that I could move it back and forth in different directions without any mechanical movement whatsoever nothing moves mechanically and when things don't move mechanically they become much faster if you deal with things that are moving electrically so once you can do that then possibly you could do a raster scanning so we made one actually a postdoc and a student on Monday ruse and through this baby actually made this and this is an this is the active part of it so this is the picture of the die this is the die fold again this is about a millimeter on the side this is very small and this is the active part of it which is even smaller so it's a four by four array it's a very small array I mean if you want to make a more focused smaller very sharp spot you have to have a larger array but we started with like a four by four sixteen to prove as a proof of concept and if the spot size would not be super small small but it would be sufficient and to demonstrate the concept so you have this that radius and this is the actual chip as a part of the test board that we have there and these things each radiates light the couple the bright light out and when what happens is that the in free space they add either coherently in certain directions or incoherently in other directions so we made this and it does bend the light what you can see is a demonstration these are basically simulations on what it should do and these are actually measurement so you can see that we can form a spot and we can move the spot left and right and up and down so you can go in both directions so you are we don't really do that so we said okay well if we can do this can we form an image well we said okay wait why don't we start with something simple like a triangle and how do we form an image if you have a spot well think about if I could move the spot very rapidly with my hand it's very difficult to make a triangle right good like hand the mechanical movement but if I have an electrical movement it's things become a lot easier so we are gonna go slow with the spot and then make it fast go faster and faster and faster and see how it performs so look at the spot this is a real measurement and we are making a triangle then we thought okay maybe we can play a little of game happy faces sad faces things of that sort so well let's see them so in the beginning you can't tell if it's doing anything but once it goes fast enough then you can see there you go right okay and then we said well we are a California Institute of Technology see if we can make the C I and T and then we made them individually said well okay why can't we just go through the sequence and that's bending the light all right changing subject so the next thing is that yes we can make these systems that are very complex but because they are made of things that are very small and there's a very large number of them the possibility of failure goes up once you've basically to think about it if you put 10 million things of anything's next to each other the possibility of at least one of them failing is quite high so to avoid that you have to go through extreme measures to make sure that everything is perfect but even in that case things are very quick on very expensive and very difficult and there's a limit to how much you can do that but now the question is that do we really need to do that or can we make circuits in such a way that they can actually heal themselves well our bodies do that right if you look at it wound it heals over time why can't I cease do that can you make chips ICS that actually heal themselves and well we went about doing this and I'm gonna not go through all the technical details so this one doesn't mean anything to you probably if you're not an electrical engineer and don't worry about it it's not supposed to so but what it is is that this is a demonstration this was a platform this is a particular amplifier but what take away from this is that this system has a lot of sensors and actuators it has mechanisms or detecting failures of different kind an actuating things changing things to make amends for that and correct them now what this does is that eventually at the end of the day it's a system it has a brain in it that basically tries to find the best it can make out of the card it's dealt so whatever it ends up with it says ok well these are the cards I'm dealt with I have a good attitude I'll try to mix it to make the best out of it right and if you do that you have essentially created a self-healing system because it does the best it can possibly do and these are examples of how it performs now as an example this is the power consumption power consumption is very important when you make electronics because that determines how long your batteries last if you have your power consumption you've doubled your battery life for a portable device so one of the things it did is that when itself healed it reduced the power consumption significantly so these are 420 different chips 20 different samples we tried this system before healing where so we had a system such that we could operate without healing and say well okay now go heal and basically when you ask it to heal it made it the power consumption substantially lower and interesting enough also things became more similar to each other it's kind of like that old adage from a bleep balls accident who said you know all happy families are the same so in a sense that you know if they are operating close to optimum they are very similar to each other and you can see that these solid lines are basically much closer to each other than these dashed lines and this is shown in this histogram so what we have here is that then we said okay well once you have something like that it should be able to correct for heart failures serious destructive catastrophic failures so you know we have these power power lasers in our lab that he can actually blast things on the chip and destroy it in different parts of the chip so we said okay it was not intended for this we said what happens let's see what if what happens if you do this so we started blasting the chip so we started blasting different parts and important parts of the circuit and see what it does and when we did that it interestingly recovered so you what you see is the basically that the red lines are after each blast what it performed before healing and the green and the blue lines are the ones that it did after two different ways of healing which essentially resulted in more or less the same performance for the most part so it's recuperated it basically reclaimed as much of the last performance as it could and it was substantial in many cases so it basically can heal itself paint knows kind of damages against those kind of damp just so we said okay well these systems can be operating in these regimes then one of the other applications where we can apply these things and again I got interested in these applications biological applications and one of the targets was really to develop this personal handheld diagnostic device a device the size of this point that you can actually use to not only perform basic tests like I'd like as glucose test or blood sugar test but you can actually use it for detecting different markers protein markers DNA RNA markers of various kinds for different applications in your blood and use it as an personal medical device personal diagnostic device so some of the requirements of these things would say basically if you want to make it low-cost and handheld and low-power it really needs to operate in certain ways and you need to use a certain kind of modality of sensing the modality we chose was a magnetic modality so basically we are using a magnetic sensing modality here which basically as an example of that this isn't an array we made of these sensor cells so one of what we see here is basically a magnified view of that and you can see basically four different sensor spots here out of the 64 that you have here and an 8 8 by 8 or 8 and each one of them well really 48 out of the 64 of them can be used for detecting different analytes so using this chip you can actually look for 48 different targets at the same time in your blood these can be protein targets or RNA or DNA targets and this is designed such that the cost structure of this is so that these are so cheap that designed to be one-time use so once you make these chips you can actually dispose of them as a part of your test so and they basically we took this sensor and integrated in the setting and we made this reusable reader so this is the reusable breather and this is our disposable makeshift disposable cartridge of course it's not going to look like this eventually but this is basically what we use it's a very simple design which allows you to detect these different targets and we demonstrated it with two different sets of targets one is a protein target basically interferon gamma which is relevant to for tuberculosis and the other one is a DNA assay and both of them basically that we demonstrated that it works over a broad range and these plots basically shows the sensitivity to these inputs so and this opens the door for a lot of other additional kinds of medical diagnostic devices that can be made and the idea would be to have a handheld device which basically you can buy a cartridge for certain kind of disease and put it in and run it and discard and then have the reader you reuse it for a diff completely different test and this would be a battery operated handheld device which would be no power enough that can operate off of a battery and can transmit the information back to your system in our design basically right now we are using a USB but this part is a standard part you can actually do it in different ways so with that in mind I like to let me bring it to a conclusion I think we we need to take a holistic approach to what we can do with our technology we have wonderful technology at our disposal we are getting to a closer to closer and closer to a point where anything that's not excluded by laws of physics can be made the question is that what are the things that you would like to make and what are the things that we should focus on now we have to really become more creative and versatile we have to learn more things about more and more about more and more to be able to do stuff and we are really limited to some extent by our imagination and our tools the ability the tools of use and we have to really expand those now we are moving toward a more closely linked world that's a fact whether or not whether we like it or not and it brings its own new challenges and opportunities but we are living in interesting times now in the end I'd like to thank the people who did the work this is really you know we as professors have the honor and privilege of working with some of the smartest people in the world and many of these people deserve some of the former and current students who have worked with me in the past many of them I have believed twelve of them are already professors at various places and many of them are running companies they have started companies and I'm really proud of their achievements but we are primarily the cheerleaders for the efforts and we really have to thank each and every one of them for the efforts that they put it we also like to thank our sponsors for the work and in the end I'd like to leave you with a thought thank you you

Channel: caltech

Views: 7,815

Rating: 4.9506173 out of 5

Keywords: Caltech, science, technology, research, electricial engineering, circuits, engineering

Id: yNTJOcM5ugs

Length: 49min 25sec (2965 seconds)

Published: Fri Aug 26 2016