Embedded - 214: Tiny Sensor Problems

Episode Date: September 7, 2017

Kristen Dorsey explained MEMS sensors: how do they work, how they are made, and what new ones we expect to see in the future. Kristen’s website is kristendorsey.com. She is a professor of engineering at Smith College and runs the MicroSmithie. MEMS stands for microelectromechanical systems (Wiki). Used in some sensors, Galistan is a room-temp liquid with interesting properties (Wiki). A few interesting MEMS applications: Micronium: a tiny resonator making music 2-stroke gas engine Pinball machine One of Kristen's stretchy strain sensor, not MEMS (so you can see it)

Welcome. This is Embedded. I am Elysia White. My co-host is Christopher White. You've heard me burble happily about inertial measurement units with tiny accelerometers and gyros. Sometimes we even say MEMS sensors. This week, Kristen Dorsey is going to help us explain all of that, how these sensors work, and how they're manufactured. Hi, Kristen. Thanks for being on the show. Hi. Thanks for having me. Could you tell us about yourself? Yeah. So I have been working in MEMS for about the last 10 years.

I'm an assistant professor at Smith College, which is a liberal arts college in Western Massachusetts. And I teach courses in circuit theory and circuit analysis and MEMS. Fantastic. So now we are going to do this lightning round thing where we ask short questions and we want short answers and if we are behaving ourselves we won't ask you for lots of detail christopher just noticed that i didn't update this list and he has to like make them up on the fly favorite book or movie uh that you encounter for the first time in last year uh oh gosh i think it's a tie between Ursula Le Guin's The Dispossessed and a graphic novel called Descender. Those are both great. Would you rather have a dinosaur or a blue whale for a pet?

Dinosaur. Preferred voltage? Five volts. How many students should be in a college classroom? I think it depends on if it's a lab class or a lecture class, but I'm going to say 18. How often should a college class meet? At least twice a week. Regular classroom or flipped classroom? Flipped classroom. Technical writing tip you think everyone should know? Document everything all the time.

As you go? Is that what you mean? Yeah. Okay. Yeah, as you go. And what would you do in life if you knew you could not fail? Fly. Like just flap and fly or like get on an airplane?

Well, if I could fall and not fail at falling or I could fail at falling, I guess, which would be flying, then I would do that. So, yes, by flapping my arms. Cool. All right. Well, let's get started with the MEMS part of the show. And while I have a lot of technical detailed questions involving diving boards, I kind of want to start out for the people who don't know what MEMS are. What's the definition? What are they? Why are they important? Right. So MEMS stands for microelectromechanical systems, M-E-M-S. And that term can be a little bit restrictive if you're holding yourself to

the electrical and mechanical part of the system. So microsystems is a slightly more encompassing term because you can have things like microfluidic devices that don't have a moving mechanical component, micro-optical mechanical system, MOEMs. And so in general, it's a term that describes systems with a feature size of 1 to 100 microns in width or length that move or have an electrical signal. And so are these usually made out of silicon, just like chips are made out of silicon? Yeah, so the original MEMS designers kind of hacked the CMOS process to create the first MEMS devices.

And so they're built out of the layers of silicon, whatever dielectric is, is separating the metal layers and the metals, just like transistors and the routing on chips are. So initially they hacked processors so that they could get sensors. Is that still how it's done? Sometimes. A lot of the commercial sensors are fabricated in similar ways, but there's been kind of this push to develop new fabrication technologies that aren't necessarily based on CMOS processes or of micro fabrication techniques to create new types of MEMS devices. But one of the advantages with the CMOS is, I mean, we're pretty good at making lots of chips fairly cheap. We are. So one of the advantages of using chip fabrication processes is that we know how they work. We can batch fabricate them. But because they're getting so small and MEMS devices in comparison are so large,

it doesn't necessarily make sense to put them on the same chip because the MEMS device takes up so much area. So then you have to decide, well, am I going to integrate the MEMS device on chip? Am I going to have two MEMS device on chip? Am I going to have two different chips with different processes? And it's still a little bit of an open question in the field, I'd say. Okay, I didn't realize they were different sizes. So like the chip circuitry, the transistors that make my chip do smart things that I tell it to do, is one size but the accelerometer gyro may be

much larger. How much larger? Orders of magnitude? Yeah, so I think commercial accelerometers in a commercial IMU unit are about 500 microns on a side, if I had to throw out a guess for order of magnitude. So that's going to take up a huge percentage of any chip if you were trying to integrate it with a lot of other things. I've always wondered why we haven't gotten, you know, the Cortex-M series integrated with accelerometers and gyros, since that's such a common add-on. And now I know. Is there a physical constraint? Because chip processes are being driven downward to 14 nanometer, 10 nanometer, and MEMS isn't following that because physics says you can't make a sensor

that small because things don't move right? Or is there some other reason? Well, definitely when you start scaling that small, the models that we have for how these devices behave get a little less effective and the devices behave a bit more strangely. But another reason is that a lot of the tricks that circuit designers have played with chip fabrication processes make it not so great for designing MEMS processes. So for example, things like the dielectrics they're using nowadays to separate the metal layers are more difficult to etch away and more difficult to use as a structural material for the MEMS devices. I think maybe we need to figure out why that's important.

So how would you build a MEMS accelerometer? It was always explained to me involving a diving board that didn't make that much sense. So hopefully you can start there. Oh, yeah. So that is one way that a MIMS accelerometer is built. But a lot of them, commonly the commercial ones nowadays, are kind of fabricated from moving capacitors. So if you think about having a capacitor and putting one end of the plate on a spring so that it can move up and down. And then, sure, okay, diving board. So if you, you can, you can squish the plates closer

together with a force or pull them apart with a force that will change the capacitance. And so a lot of these moving capacitors, the moving side of the plate are attached to a mass, a big structural just mass. And so you can get a shift in capacitance that's very large for an applied force or applied acceleration. Okay, so we have an accelerometer and we have a plate. And we have some space in between and another plate. Because that's kind of the definition of a capacitor.

The plate and then some stuff that doesn't pass electrons easily. And then another plate. And because one of these plates is on a spring, like a diving board is on a spring, it can move closer or further apart, and thereby changing the capacitance of this whole little capacitor module. Right. Only instead of the end of the plate being on a spring like a diving board where the tip is moving up and down, the whole plate moves back and forth. So like the middle of the spring, the middle of the plate is connected to a spring that's then fixed somewhere else.

Okay. So the whole plate is moving closer or further away from the fixed plate. That makes sense. And it's moving closer or further away according to an acceleration, an outside force. Right. And so this only measures one degree. And we always get three axis accelerometers now. But that just means that there are three of these in there. You can have three of them in there. You can measure the change. You can kind of play special tricks so that you design the spring in such a way that you get a change in capacitance one way if there's an acceleration in Z, like up and down, versus if there's an acceleration in the plane of the accelerometer.

So I don't actually know too much about how commercial accelerometers measure in all three axes. They're a bit secretive about that. Now I have a metal box hung by multiple springs in my head. So we'll just imagine that to be the truth. What about gyros? The weird thing about this explanation is it always started out with diving boards and then ended up with Coriolis effect and frantic hand waving. So maybe you can explain how MEMS gyros work. All right. yeah. So let's go back to our accelerometer because that's what the gyro is built on. So we have a mass with a bunch of capacitor plates attached to it, and that mass is on a spring and it responds to acceleration.

And so when we accelerate or put a force on that mass, the capacitance between the moving fingers and the fixed fingers changes. So a gyro is like that with one more layer of capacitor fingers and another mass. And so the inside mass is moving. It's stimulated by an electric field. And so there's an electrostatic force that's moving the mass to create a velocity. And then when the device turns, this is where the Coriolis force comes in, when the device turns, a force is generated that changes the capacitance on the outside set of fingers. So it is actually kind of close to the mental model of a metal box with a bunch of springs

coming out of it. That is what our models look like when we draw it. Okay. out of it. That is what our models look like when we draw it. Okay, Chris is nodding, and I'm still trying to figure out how we got from an accelerometer, which measures acceleration, to another layer of things that ended up now measuring spinning. Maybe, Chris, you explain it it you were nodding well it's a bunch of plates that twist right instead of coming moving closer oh is this like having polarized sunglasses where you if you have them in line you can see through and if you have them out of line that you can't see through is it sort of like the capacitor version of that? A little bit. So if you think about if you throw a ball, if you try to throw it towards another person, if you're more talented at sports than I am, you'll probably throw it to them.

But if you're on a rotating table while you do that and you throw it, where you think the ball will end up and where it ends up are kind of mismatched. So that's what the Coriolis force is explaining, is that difference between how a ball travels when you're standing still versus how the ball travels when you're rotating. And so the gyro measures that kind of difference between non-rotating motion and rotating motion by a second set of capacitors that are changing as the rotation is applied. All right. I can buy this.

I don't think I could explain it to someone else without reverting back to hand-waving, but for this moment, I'm good. Okay. There are lots of MEMS sensors, not just accelerometers and gyros. Which are your favorites? I think my favorites are they're the type of sensors that I work on now so sensors that are made out of like rubber or silicone material for sensing forces or strains or tactile sensing so the the grand vision of these is to make an electronic skin for things like VR applications and robotics. I have so many questions, I don't even know where to start. Okay, first, rubber is not CMOS, so these must be created with some other methodology. Maybe tell me about that. Yeah. So the process that's used to create these

sorts of devices looks a lot more like newspaper printing than it does like making a transistor. And so there's lots of different ways to make these sensors that are fabricated in rubber. But one of them is to apply a rubber layer by spinning it on and then put down a metal layer. And there's lots of different ways to do this. One of my favorites actually uses 3D printing. And then you put down another rubber layer by spinning it or however you like. But these are still MEMs, so they're still tiny, tiny things, right? Yep, they're still ones or tens of microns big. Okay, so now you have these itty-bitty little sensors

that will tell you, or what do they output? Do they just output a current or a voltage when they're stretched or not stretched? So most of them, they look like resistors or capacitors. So just like a wandering line, if you've got a resistor. And then when you stretch it or apply a pressure or twist it or anything like that, the resistance changes because you're changing the length or the cross section of it. And so the change in resistance is the sensor output. And so if I had a glove with one of these on each joint, or maybe 10 of these on each joint,

when I closed my hand in VR, the virtual reality hand would close because it would be able to tell all of the strain and pressure points. That's the hope. Okay, when can I buy that? I think it's unfortunately still a ways off. One of the sad things about MEMS has been that I keep hearing about these fantastic sensors. I remember hearing about chemical sensing in like 2001 and how that was going to change the world. MEMS chemical sensors were just going to be fantastic.

And it never got here. I still don't feel like the chemical sensing is here. Are these really going to happen? Are you just taunting me with futuristic technology I can't have? I don't want to say, no, we're never going to have like a chip on a nose, a nose on a chip kind of chemical sensors, but I think it's really a really, really tough thing to do. So we've got all of these different ways to make chemical sensors and they're very sensitive and they can measure parts per million or parts per billion. So, you know, one particle in a room level of chemicals. But one challenge is that they're not specific. So they might respond to methane and be able to sense that, but they might also

respond to something else that you don't want to sense. And so that's really the challenge with chemical sensors is how do we get them to measure things that we care about and ignore things that we don't care about? I understand that. Our smoke alarm recently went off because it had too much dust in it. And I really thought it should be able to tell the difference between dust and smoke. Yep. And so even at a very small scale, you can't specify the chemical you really want without allowing false positives. Yeah, it's a really challenging problem. Do you have similar problems with strain and pressure sensors?

Right. So these sensors change with strain. pressure sensors? Right. So these sensors change with strain, they change with pressure. They'll also change with temperature. Some of them change with humidity. And so figuring out ways to separate out all those different changes is a problem. All MEMS sensors are temperature sensors, right? Because they're so tiny, they have no choice. They are all temperature sensors, right? Because they're so tiny, they have no choice. They are all temperature sensors, yes. And so an accelerometer in gyroland, usually you can calibrate that out or make some,

even if you aren't doing per-unit calibration, you can make some system generalities related to calibration and taking into account temperature sensors. Many of the IMUs that are out there have temperature on board so that the sensors can be fixed. Right. Do you have the same linear or easily binomial solution to temperature? Or does the rubber make it harder? The rubber makes it a little bit harder.

But I think as a field, we're kind of at a level where we're still trying to figure out how to make these sorts of sensors really, really sensitive and how to put a bunch of them in a glove at the same time. And I'm not sure that many people are even tackling the problem of, well, how do they change with temperature yet? Okay, what problems are you tackling now? Just making them? Yeah, just making them. So what's the difficult thing there? One of them is the fabrication processes make it difficult to, the variations that you get from device to device make it really difficult to have two devices that behave exactly the same way.

Ah, yes. Scaling up for manufacturing is always a difficult thing. And I imagine at tiny, tiny size, it's even more difficult. Tiny, tiny size with new processes. Yeah. Another one is in figuring out how to make these applicable to things like VR. So if you, I really liked your analogy about putting these in a glove and then closing or opening your hand in virtual reality. So if you have a bunch of these sensors, you've got them on each joint and you're closing and opening your hand, you'd be able to tell how your hand is moving. But as you move your hand around, maybe those sensors migrate away from your joint a little bit, and so they're not measuring exactly the same area of your hand. And so that's one way that you get kind of error in your system over time.

And that's a common sort of error. I push everything back to IMUs because it's what I'm very useful in. Definitely, if you have a measurement unit that isn't in the place you want to measure, you have to start doing some interesting trigonometry to pretend you are in that location. Right. But that has a lot more information than just, is there a pressure around me? Those have positional information and having a migrating sensor in a glove, you wouldn't be able to, you'd have to have a lot of them. Right. and once you have a lot of them then it's um then it's a challenge of well how do you how do you deal with all of those signals at the same time how do you get them off the glove and into processing are you going to make someone wear a bunch of wires around for the prototype yeah you said that making them is is challenging and definitely you mentioned 3D printing, so that seems dicey.

What kind of 3D printer can work at these scales? A lot of people are building their own. Yeah. you know, custom built everything to try and work at these scales and get low variation at these scales. But then you run into micro fluidics issues where you have to put down very tiny fluids without having to deal with all of their weird chemistry behaviors when they're small. Yes. So I guess the 3D printed devices that I've seen are more like 100 microns feature size than 10 or 1 micron feature size, but still pretty small, pretty challenging. Would milling them be easier? There are some people that work on milling, micro-milling processes. I think though that because usually we're, for these sorts of sensors, materials like rubber and also this cool material called E-gain, eutectic gallium indium, which is a gallium indium mixture that's liquid at room temperature like mercury is.

So it's difficult to mill things like that. So are we back to trying to have solid things and chemical etching all the CMOS processes? For these devices, for the flexible microsensors, it seems like more people are taking the approach of printing or depositing, using some sort of additive manufacturing on a piece of rubber rather than trying to etch things away.

Yeah, because rubber interacts with everything. Yeah. Okay, so these don't sound manufacturable. What else do we have coming down the line? They don't sound manufacturable right now. Eventually, 3D printing at that scale will get so much better and it will be a revolution

on lots of things. But right now, what other sensors should I be excited about? So I think one thing you can be excited about with chemical sensors is that the machine learning approaches to figuring out what chemicals you've got

are getting much better and processing is getting much better that way. So we may never be able to have a chip that acts like an electronic nose, but we may be able to tell much better than today what sort of chemicals are in food or in the air, that sort of thing. Oh, my beagle is pleased that she will not be being replaced right away. What about MEMS actuators? I mean, how can something that small get anything done? So MEMS actuators get used a lot in service of something else. So MEMS resonators for tuning or filtering or switches for like MEMS mechanical switches instead of transistors. So the actuation isn't, it's not a big

actuation like you might imagine at larger scales. Yeah, I'm imagining a giant robot arm. So yeah, maybe not. Right. So one place that MEMS actuators do get used is in micro robotics, micro scale robotics and so outputting very small motions in in a fluid or on a table to propel these robots someplace else okay nothing bad can happen with this I have never seen a movie in which these robots go insane and kill people but But are we really going to build tiny robots we can't even see? I think, yeah, that's been a long-held dream for a lot of people. One device that I'm equal parts in awe of and terrified of is the idea of taking something like a pill camera that you swallow that has cilia on it so it can propel itself through your digestive tract and check everything out.

And I think this is a really cool idea, but also has a lot of bad sci-fi movie tropes there. But this is, I mean, putting MEMS little robots in our bodies is actually a fantastic idea once you get away from the sci-fi tropes. Because controlling bacteria and the bioengineering side that way has those chemical downsides. But if we make it robotic where it runs out of energy at some point, maybe it goes away. I mean, Chris, you're looking at me like I'm crazy. Am I just exchanging one sci-fi nightmare for another? I'm not sure. Those are some pretty small robots. are some i mean little little legs pushing around a camera that's fantastic that's bigger than a bacteria though it is thinking that

there's a little there's a gap between the camera thing and going around targeting cells how big is that gap we're all gonna have to google how big bacteria are right now, aren't we? That's all frantically go to our computers. I'm sorry, Kristen. We don't need to write the horror movie right now. Okay. But I said about them needing power, and there are MEMS turbines. I mean, did I really read that?

Can we do power generation? Yeah. So there are MEMS turbines and energy harvesters. It's kind of tough to get a lot of power out of them right now. But the idea for most of those is that there's some sort of vibration, like from someone walking, that gets transmitted, either through piezoelectric material that transduces pressure into an electric field, or through something like, again, the capacitor that transduces that into an electric field to store that energy. That's true. If accelerometers are capacitors, then we're generating energy every time we jump up and down.

Mm-hmm. So this goes back to chemical sensors. I've been hearing about this for a while. When are we going to have it on our desktop? The chemical sensors? No, the power generation through piezo vibration and movement. Ah, so I think the MEMS energy harvesters, again, are one of those things that every year people are like, oh, this is a year out. And I

think they're doing a really good job of advancing how much energy they're able to harvest. So now I think the limit is about 10 to 100 milliwatts continuous. So when someone is walking, they're generating about 10 to 100 milliwatts at a time. So it might not be enough to power a wearable device or charge a cell phone all the time, but providing a fair amount of

energy. That's enough power to keep things that are normally asleep powered. Yeah. And things with displays couldn't get powered on, but it's starting to be enough power to power the accelerometers and the gyros. Yeah. And things with displays couldn't get powered on, but it's starting to be enough power to power the accelerometers and the gyros. Yeah. And then you just need a battery for your processor and off you go. So one of the places I think that energy harvesting is going to get to be really useful is in monitoring things like bridges. So when you have a continuous amount of traffic going over the

bridge, generating those vibrations, and then the system can wake up maybe once a day and say, hey, I sense that there's cracks here, or hey, everything is okay. So that you don't have to go out and replace those batteries every two years or five years or however long. Yeah, I've sat in on that meeting for a few companies. I'm looking forward to when we have those. But I admit that, again, around 2000, 2002, we were talking about how to do this with MEMS Power Generation and IMU and Tiny Radios. And happily, at that time, came up with the idea that it wasn't quite possible.

But I am looking forward to it for the future. Yeah. Some people have made relatively silly MEMS sensors on the idea that if you can make it MEMS, you should make it MEMS. Chris came across a two-stroke gas engine. Had you seen that before? I hadn't. That was fantastic.

I'm not sure how small it is. It wasn't clear from the article, but... But it said MEM, so it must be small. What are the silliest things you've heard of? Or maybe done? I haven't done too many silly things, sadly, but the two most fun devices that I've seen, a couple of years ago, someone made a MEMS pinball machine. and controlled it using, it was an electrostatic force for the little flipper paddles,

and it would ping all over the screen. That was really cool. How big was that? I think the whole thing was about 250 microns by 250 microns. That's amazing. This is one of those areas where I just don't, you know, it gets beyond a micron and I'm like, I don't even really know what that is.

I know what it is when you have a whole bunch of them together, you end up with a meter eventually, but at some point it's, it isn't believable. I mean, it just doesn't sound right. Okay, so what other sensors or silly sensors have you heard of or things you look forward to seeing? Well, my other favorite fun sensor is the one that I show all my students at the beginning of my MEMS class. It's the world's tiniest resonator. And I don't know whether or not it plays the world's saddest song. Okay, a resonator.

How do you make a MEMS resonator? We have the piezo crystals resonators. Aren't they the same? Right, so that's one type you can also again going back to the the the moving mass with the uh capacitive fingers on it so instead of so you've got a mass and a spring uh together and that that makes a resonant frequency and then you drive it with by applying an electric field between the moving capacitor finger and the fixed capacitor finger okay that makes sense sort of um why is that i mean okay so you can make it play a song

why is that one your favorite? Is it because it's both simple and can lead to more complex behavior, or is there more behind it? The mass spring comb finger arrangement is one that gets seen all the time in MEMS. I mean, I've talked about it a lot, but it gets used for chemical sensors and accelerometers and resonators. And so I think showing students that, yeah, this can make noise if you have a big amplifier. But also, you know, this is kind of the building block for everything we're going to talk about in this class really brings it home in the first week for them neat what else is your research about i mean when we talk about mems that's big and there are lots of industrial applications and industrial versions but you work on specific stuff what What do you work on? So I work on the flexible strain and pressure sensors that I was talking about earlier.

And my specific focus is in figuring out how to place cuts or empty regions where there's no rubber around the sensor to change how it behaves with strain or pressure. So maybe you want to use this sort of sensor to sense strain or pressure, but maybe also you want a temperature sensor on board that you don't want to respond to strain or pressure. And so by kind of cutting out the region around the sensor, you can change how it behaves with strain and change its sensitivity. Since you're making these things out of rubber materials, what kind of...

It seems like when I think of electronics, I think of it as very stable and doesn't wear out and doesn't, you know, I guess could corrode. But in rubber, I think of as rotting away sometimes if it's in the wrong environmental condition. So you probably have a lot of different environmental testing kinds of things you have to do versus an accelerometer. Is that true or yeah so um one one problem we have in the lab is um that they do change a lot with humidity and so

we have we have to put it in a in a dry box so that on a rainy day it doesn't behave differently than on a sunny one. But I think we're thinking more of the realm of single-use or almost disposable devices rather than something that's going to be around for a long period of time. How do you identify how these things behave when you have some trouble making them consistent when you produce them? Or did I just sign up for a PhD paper? Well, I think that it's definitely a challenge to figure out an exact model

for how they'll behave when we haven't built them yet and when there's manufacturing variations. But I hope as our fabrication process gets more dialed in, the sensors will get more consistent, more uniform, and then we can characterize them. But as to how we characterize them, we've got a bench vise that's attached to a servo motor that applies a particular strain, and then we're measuring the capacitance or the resistance of the device for different strains that we're applying. You said earlier it was mostly about the resistance for these devices. Do you also measure the capacitance to see if there is an orthogonal thing to look at,

or is it just part of the testing setup, or is there more to it? And I didn't hear about capacitance earlier. Ah, so some devices are resistive devices, and so the resistance changes. Some devices are capacitive, so the capacitance changes. And it kind of depends on what application, how fast you want them to be, whether or not you use resistors or capacitors. Well, it makes sense to measure both at these scales. There's a possibility you have something you didn't know you had.

Okay. I mean, this is sort of like the apocryphal story of the guy who invented saccharin because he was too stupid to wash his hands after dealing with chemicals and then noticed his lunch was very sweet. And happened not to die immediately. Yeah. and happen not to die immediately but you know you might as well test everything as long as the testing's there so why why does

stability matter? That's what my outline says well let's see, why does stability matter? So we can reproduce things? but how do you debug with this? I mean, they're so tiny. You can't just go in there and check each point in a circuit because you can't touch them. One way that we debug is by looking at it. So using an optical microscope or scanning electron microscope, SEM, to see if there's some obvious physical flaw. Like, oh, that's touching that other thing.

That's why that's not working. We also use just a lot of electrical characterization to figure out, you know, did the resistance or the capacitance shift a lot? Yeah, that device probably broke. But this working in MEMS is kind of, especially newer MEMS areas, is kind of build a lot of devices and hope a fair number of them work. And the yield, the number of successful devices that you get is just not super high. No, not until you have it dialed in. It's a very chicken and egg problem.

Yes. How well do your models work? I mean, you mentioned that you do put them in a model, and so I imagine you do some simulation, physical simulation, so you expect things before you print them. Are they close or are you finding all of the ways they're not close, like humidity? They are, they're close and not close. So I use a computer program that does finite element modeling of these devices. So I build a model of it, a physical model, and then look at the physics of each part of it. And that gives me

enough information to say, okay, well, the response should be about this much. But it doesn't catch everything. So for example, one thing I didn't take into account in my last model was that these things are made out of a sheet of silicone rubber, and it's thin. So it sagged a little bit when I was testing it, and that totally changed the response of the sensors um from when i i had it completely flat and taut and was was testing it just that that extra force of gravity and so i wasn't testing that in my model and so i was like

okay i got to go back and and add that in so it's kind of an iterative process, starting with the model, building it, seeing how it works, and then going back to the model and adding new physics in. Adding new physics in. I like that. Having to have gravity be a factor makes sense because these sensors are so tiny. The accelerometers, they measure gravity, but it's just a force to them. It's not like they are affected much by it. So I understand how you didn't initially perceive that gravity was going to be important on something this small. Are there other forces you have to worry about? So I was working on a project a couple of years ago with chemical sensors called, they're metal oxide sensors that respond

to a bunch of different chemicals. They change their resistance. And they're most sensitive when you heat them up to about 350 or 400 degrees C. And I went outside and I tested some of these and I noticed every time a truck went by that my readings would go totally haywire. And I thought it was maybe because of the truck submissions. And it turns out that because these are heated to a really high temperature that any change in temperature really changes their resistance. So anytime a wind gust blew, the sensors would cool off for a little bit and totally change the output I was getting. So in that case, it was wind and on one hand that makes them impossible to use on the other hand for a long time they said that cd drives would be impossible to build because you couldn't have

that level of precision and accuracy and robustness and now that's a done technology that we aren't even bothering with anymore so So I think we will get to MEMS that we can use. And I think you're right that machine learning is going to be part of it, because you're going to have to have extra sensors to figure out things like that. I think so. And so the cool thing I found out with those particular sensors is that by using just proportional control on the heater resistance and keeping the heater resistance constant kept the sensor output constant because it would change temperature to respond to things like

wind. Yeah. It's a solvable problem. Once you know, it's a problem. Yep. Plus you've invented a wind detector. Searching around for MEMS gadgetry,

I did come up with a MEMS windmill. I thought that was really cute. It was very tiny. They put it on a penny. Accelerometers and gyros are very susceptible to vibration. Somewhat surprisingly susceptible to vibration. If you need a vibration sensor, you can just get some of these. Is there other stuff? I mean, do other sensors have to worry about vibration? Is that a,

like temperature, just a tiny sensor problem, or is it specific to the inertials? It's specific to having the type of MEMS sensor where you've got a mass suspended by a spring. So you see that sort of MEMS sensor type in most accelerometers and gyros, but also occasionally in chemical sensors. So then those chemical sensors would be sensitive to vibration. Can't think of any other ones off the top of my head. Okay, cool. Let's see, I have just a few more questions. You have a PhD, because that's kind of professors usually do. And how did you decide whether to go into industry or academics? I was dead set on going to industry until I actually had an internship. And I realized the favorite part of the internship was working with other interns to do tutorials for particular things. Like, here's how to use this program, or here's how an accelerometer

works. Here's the basic equations. And so once I got back from my internship, I was like, I really enjoyed that teaching part. Maybe I should figure out some way to do that. Fair enough. And the research part? So that's the cool thing about my job at Smith College is that it's equal parts teaching and research. So I get to play with enough zany ideas and try out new things in the lab. But then I also get to spend a lot of time with my students, both in the classroom and in the lab,

working with teaching them. And what advice do you have for, say, college sort of problems to work on and what area to go into. So you're a proponent of the think ahead for what you want to work on. Don't just end up in a program and work on whatever your advisor works on. Yes. That makes sense. It does seem like there are two paths because sometimes you end up at a school for reasons other than they have the program you want. Maybe they have a program that was acceptable in an area that was good for a price that was reasonable. And it's hard to chase the,

I want to work on robotic prosthetic limbs without going to Canada. Great. Sorry, there was a past guest who talked about that. I was fascinated. All right. So one more question before I ask Christopher if he has any other questions. What areas of technology do you think will be most interesting in five years? really informed by working at a liberal arts college. But I think the coolest projects come out of fusions of multiple disciplines. And so, you know, five years, 10 years from now, I'm not sure who's going to decide over

launch, hey, maybe we should, you know, try and mix psychology and bioengineering to figure out this problem. So it sounds like we should be having lunch with lots of interesting people. That's always my other good advice, yeah. Chris, do you have any other questions? Well, I think we've talked about it around the edges, but I was wondering if you had any advice or if you had things that you wished end-user engineers understood or knew about these sensors or appreciated? Because we kind of tend to think of them as, oh, it's a magic thing. I put it in and I get these numbers out and everything works fine. But are there any frustrations you have with the way people use them or think about them or you wish they knew? I think that, so this again goes to something that we've been talking about kind of throughout the podcast, but I think there's a lot of hype about what these sensors can do in the lab once or might be able to do in five years. And people understandably get

really excited about, hey, I heard about this sensor or this actuator that can do X. And so I guess a bit of, I wish more people had a bit of skepticism about the capabilities of these sort of sensors just so they don't get burned out about, you know, I heard that 20 years ago. Yes, chemical sensors, energy generation, okay. Yes. But, I mean, a lot of new technologies have this overhyped face. Batteries. You see these papers about, oh, I have a new battery.

It weighs a pound. It can store, you know, 700 kilowatt hours, but it only works once. Yeah. So, yeah, the skepticism is healthy, I think. Fusion. Right. Any day now we'll have fusion. All right. Well, Kristen, do you have any thoughts you'd like to leave us with? My thought I'd like to leave you with is to go out and have fun with projects. All right. And I think that's great because there are lots of fun projects out there. Our guest has been Kristen Dorsey, Assistant Professor of Engineering at Smith College.

Thank you for being with us, Kristen. Thank you. Thank you to Christopher for producing and co-hosting special. Thank you to our Patreon supporters for Kristen's microphone. You make the show better and my life easier. And of course, thank you for listening. I have a quote to leave you with this one from Roald Dahl. And above all, watch with glittering eyes the whole world around you, because the greatest secrets are always hidden in the most unlikely places. Those who don't believe in magic will never find it. Embedded is an independently produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California. If there are advertisements

in the show, we did not put them there and do not receive money from them. At this time, our sponsors are Logical Elegance and listeners like you.