[From the last episode: The big change in sensorsA device that can measure something about its environment. Examples are movement, light, color, moisture, pressure, and many more. is scale. We tend to refer to scale using “macro” for big or metric prefixes like “micro” and “nano.”]

Now that we have some semiconductorA material that, under some circumstances, can conduct electricity and, in other circumstances, cannot. basics behind us, as well as a sense of scale, let’s look at one of the big technology advances that has revolutionized sensors. It’s called MEMS, which stands for Micro-ElectroMechanical System. This name brings several components:

• Micro: the name has the scale built in. We’re creating things at the micro scale, with features measured in microns.
• Electro-: these devices aren’t electronic in the sense that regular computer chipsAn electronic device made on a piece of silicon. These days, it could also involve a mechanical chip, but, to the outside world, everything looks electronic. The chip is usually in some kind of package; that package might contain multiple chips. "Integrated circuit," and "IC" mean the same thing, but refer only to electronic chips, not mechanical chips. are, but they use electricity both to make things happen (actuate) and to sense something.
• Mechanical: this is the big deal: we can make things with moving parts at this incredibly tiny scale.
• SystemThis is a very generic term for any collection of components that, all together, can do something. Systems can be built from subsystems. Examples are your cell phone; your computer; the radio in your car; anything that seems like a "whole.": the finished units can be pretty complex, involving multiple moving parts as well as circuits for “cleaning up” the sensed signal. We’ll talk more about this in a minute.

MEMS technology has been around for quite a long time, and you can find super-cool images on the internet. I’d paste some here, but it’s hard to find the really cool ones that aren’t copyrighted, so I’ll simply leave this Google search link here for you to browse.

There is also a thing called NEMS. If you think about the scale thing, you can probably guess what’s different: micro becomes nano. In other words, they’re like MEMS devices, only with features measured in nanometers (billionths of a meter) rather than microns. Not much of this is happening yet – you find it mostly in fine wire filaments that are nanometers wide in a system that is otherwise micron-scaled. So, today, it’s mostly MEMS, perhaps with a little NEMS alongside.

MEMS Reality: Research vs. Commercial

If you looked at those images, you saw some pretty cool structures. It’s amazing that we can build them on a waferIn the context of making circuits, sensors, and actuators, a thin, round slice of pure silicon. Multiple devices will be made on it; it will then be sliced up to separate the individual chips. (and we’ll look at how over the next couple weeks). But MEMS technology is surprisingly old – researchers have been working on this stuff for many years. So why all the excitement now?

It’s one thing to make a MEMS die in a university lab, where you prove that, in principle, it’s possible. It’s quite another to turn it into a business. A business needs profit, and that can be a tall order for new technologies.

We’ve seen wafers and the many dice that can be placed on them. Depending on the wafer size and die size, you might have 100 dice or 100,000 dice on a wafer. Let’s be clear here: a wafer is an expensive thing to build. I’ll make up some numbers (since the real numbers can vary widely), but if it costs $5000 to make a wafer, then whatever you get from the wafer must sell in total for more than $5000 to have a chance at making a profit.

Wafer processing isn’t perfect. Little things can go wrong: dust particles can land on a wafer and block any light, screwing up the photolithographyA way of creating patterns on a silicon wafer for selective deposition or etching. You can think of it as taking a picture of the pattern and having it show up on the wafer. In the industry, photolithography is often abbreviated as lithography.. Materials don’t always deposit perfectly, and etch can have occasional glitches. So there will always be some dice that end up not working.

The dice on a wafer are tested when the wafer is done. The testers use tiny probes to see which dice work and which ones don’t; they then mark the bad ones somehow to identify them. When the wafer is diced up to separate the individual dice, the bad ones are discarded so that no more money is spent putting them into packages.

Getting Better Yield

There’s an important yield metric in the semiconductor industry that measures how good your processing is: net dice per wafer, or NDW. It can be a number (say, 750 out of 1000) or, more generally, a percentage (75%). The lower the number, the more you end up throwing away, and the fewer dice you can sell. In the limit, if you spend $5000 to make a wafer and you get only one good die, then that die has to sell for more than $5000 in order to make a profit. Not many dice can command that kind of price. Raise that number from 1 to 100, and now the price of each one has to be over $50 to be profitable. NDW is all about profit.

So yield enhancement is an important function, and many engineers spend long hours looking at dice that failed and figuring out how to improve the processing so that they don’t fail that way in the future. Notably, with MEMS, the more complicated things are, the harder it is to get them to yield well. That’s why many of those super-cool MEMS dice aren’t going into real products today.

So, while MEMS has been around for a while, what’s changed in the last decade is that folks figured out how to build MEMS dice that can yield well enough to build a business. No, they’re not going to have a Swiss-watch-looking mechanism in them (although they’re still complex), but they do a useful job that someone will pay money for.

Cleaning Up the Signal

One of the other important things to remember is that, in many cases, a MEMS version of a sensor may not be as accurate as an old-school macro version. Big gyroscopesA sensor that detects when it changes direction. are more accurate than MEMS ones, but you can’t fit them into a cell phone. That said, the MEMS ones are good enough to be really useful, and you can often compensate for their weaknesses.

One way of doing that is to clean up, or condition, the sense signal before sending it out. Here’s an example: let’s say you want to build your own gas gauge for your car, so you put a really sensitive sensor in the gas tank to see how full it is. Well, it’s not going to work well, because, when you move, the gas is going to slosh around and the signal is going to be bouncing so much you can’t really tell how much gas there is. You need to dampen the sensor or average the data or do something to that signal to get rid of all that sloshy noise. Same thing with sensors. That’s why conditioning is typically part of the “system” part of “MEMS.”

One final note: not all miniaturized sensors use MEMS technology. But it’s a big enough story for us to spend some time seeing how it works.

Next we’ll go through a made-up example that illustrates how we can build these things.