Wednesday, April 7, 2010

SparkFun SEN09423 integration issues

Anyone seeking to use SparkFun's SEN09423 breakout board for the LPY530AL as a position sensor should be advised that the two 4.7 µF capacitors (C1 and C2 on the schematic) used for the high pass filter need to be removed and the contacts bridged. This image shows which tiny bits are of concern, however note that it seems the resistors indicated therein do not need to be removed. This information comes thanks to a few people who know what they're doing (which excludes myself), as discussed on the SparkFun forums here and here. From what I gather this may be an issue with numerous (all?) SparkFun breakouts including ST rate gyros, the two threads alone implicate boards containing LPR530AL or LPY530AL, including the IMU 6DOF Razor. This is a particularly odd case because Inertial Measurement Units are mostly used for dead-reckoning, and the inclusion of these caps will effectively frustrate anyone with such an intent. As far as removing them, good luck! Here's my own picture of how gigantic these caps are:

I found the best luck (given a fine tip soldering iron) with adding a little solder to one side so that solder wick can get most of it. Then just heat up the other side and push gently. The first one I removed took the contact pad with it, if that happens to you you may or may not be high and dry. I managed to salvage the situation by drawing between the appropriate areas with a pencil. In case you weren't aware, graphite is conductive--clearly this is a handy bit of information on occasion.

For a slightly more general audience, here's some interesting information. The capacitors pictured are about 0.065 inches wide, or 1.66 mm; the skinny dimension of the penny pictured is about 1.52 mm. I said these capacitors are gigantic, and relatively speaking this is true! Relative to molecules, light rays, and subatomic particles sure, but also relative to the vast majority of capacitors out there. We will get to how in a minute, but first a brief overview. The electronic components most of us are used to seeing are the ones attached to those (usually) green boards also known as circuit boards, like this one:

These days most circuit boards we encounter are printed circuit boards or PCBs, called such because the production process resembles printing to varying degrees. The principle elements of a PCB are, put simply, fiberglass, copper or other conductive metal, and solder mask. The fiberglass makes up the board-ness, the copper is akin to wiring for conducting electricity amongst the components, and the solder mask, the colored part, is a coating that solder doesn't stick to, in place so that connections aren't made accidentally by wandering solder. Not too long ago, I thought the PCB was made of silicon; after all, electronics are associated with silicon, and from a naive perspective the shiny green board looks like something that might be called silicon. But if that's not it, where's the silicon? In an IC of course! These days most all the action of an electronic device happens in an Integrated Circuit, which looks something like this:

Inside that chunk of plastic there's a wafer of silicon, which could contain anywhere from hundreds to Billions of electronic components. Wouldn't it be nice if there was a window that showed the silicon? Like this one?

Instead of discrete components like the capacitors I shared above, these components are formed by spraying (very precisely) successive layers of various chemicals in a process called photolithography, resulting in something like a miniature PCB. The CPU is the biggest, most complicated IC in the box that is your computer (unless you have a very fancy video card), and because of this it looks different than all the others. For one, you can't even see it, it's hidden underneath a big heatsink, which is there to help get rid of all the electricity that turns into heat in the CPU (the process is conceptually similar to heat generated from friction). CPUs generate so much heat that one would burn itself to a crisp almost instantly without a heatsink. But even if you remove the heatsink (after you've turned off the computer), modern processors have another metal plate which hides another sealed package that finally contains the silicon. Here we're finally at the land of magic: as of now, April 2010, Intel has a 32 nm manufacturing process, which means that the typical component width is less than 32 nm. This also means that the 1.66 mm wide capacitor above is about 52,000 times wider than a single component on a 2010 Intel CPU, or, relatively gigantic. Granted, most things we know are relatively gigantic compared to 32 nm, particularly since that's quite a bit smaller than the shortest wavelength of visible light--violet, at 400 nm. Reality check: we're making electrical components so small that a ray of light can't even hit them, so small that even the most powerful microscope couldn't see them, way smaller than the average bacteria. Really!? Apparently that's not enough, industry projections have us with 11 nm chips in 2022, which would make each component about the same width as 55 carbon atoms. Interestingly, the first time a single carbon atom was photographed (after a manner) was 9/2009. Of course, there are certain problems that what we know as computers, that is Turing class machines, can't solve--certain problems that could be described in a hundred or so lines of computer code that would take a computer the size of the universe longer than the universe is supposed to exist to solve. Not content to take limitations as they're handed to us, work is well under way to develop a different class of computer: the quantum computer. Quantum computers are very different in that they can take very specific problems, like the one I just mentioned, and solve them instantly. I don't know enough about quantum computation to judge if they'll ever reach the ubiquity our Turing machines have, but I can say one thing for certain: there's not much certainty in the future! Intel will probably plug ahead and reach 11 nm in 2022, but the real question is will that even be relevant? I'm willing to bet not, it almost seems like sitting in 2002 and projecting that by 2012 our CPUs will run at 11 GHz; as it turns out, GHz aren't all that important. Take a top of the line 3.8 GHz Pentium 4 from 2004 and I assure you a 1.8 GHz chip from today will outperform it. Maybe the state of the art in 2022 will be a 100 MHz chip with a million cores--only time will tell.

No comments: