This week, among many other things, I’m working on putting together a new sensor board that I designed. The intent is for it to be used in an experimental study next week, so there really isn’t time to have them sent out for assembly.
Here are some of the parts in their bins, ready to get placed on the boards. For a scale reference, the smaller bins might just barely fit a quarter. Some of the parts are pretty tiny (.04 inches by .02 inches.)
There are a couple of different methods for putting boards like this together. It’s possible to do with just a good pair of tweezers, a soldering iron, and solder, but it takes a good eye, a steady hand, and a lot of time.
Rather than soldering individual parts one at a time, something called solder paste is used. Solder is a lead/tin alloy (well, at least the non-environmentally-friendly kind that’s easier to work with) and, of course, is a solid metal at room temperature. Solder “paste” is actually a suspension of a bunch of microscopic balls of solder in a thick flux (flux is a chemical that helps the solder flow, and as a detergent to clean the surfaces being joined.)
On a large scale, solder paste is applied to boards by using a laser-cut stencil, with openings corresponding to the connection pads on the board. The stencil is placed on top of the board and a squeegee is used to pull the paste over the stencil, very similar to how screen printing works.
I wasn’t thinking ahead far enough when I ordered the boards and didn’t get a stencil for them. So, the alternative is to dispense the paste manually. The paste comes in a syringe, so this basically consists of squeezing out the right amount of paste onto each pad on the board. This is made much less tedious by a machine called a paste dispenser, which basically connects a compressed air hose to the back of the syringe instead of the plunger, and a foot pedal activates the air for just the right amount of time to dispense the right amount of paste for one pad. It can also be set to run on a timer, so that it will continuously dispense pad-sized amounts of paste with a short delay in between each, without having to re-press the pedal.
Once the paste is on the boards, then the components need to be placed. Again, there are a few ways that this can be done. For just a few boards, provided one has a steady hand (i.e. hasn’t had too many cups of coffee) it’s often fastest just to use a fine pair of tweezers. For larger numbers of boards, a manual pick-and-place machine helps speed things up (this is the technique I’m using.) This is a tabletop device with a carriage assembly that’s moved by hand. The carriage can move along the X and Y axes, up and down a small amount, and can rotate the part that’s currently being picked. The boards go on the surface of the table (held in place by magnetic brackets) and the bins of components are at the side of the table. The business end of the carriage is a hollow bore needle pointed downward at the table. When the needle touches something, the machine starts suction through the needle. When it touches something again, the suction stops. So, the basic routine is to touch the part to be picked, causing the suction to keep it held to the needle, then slide the carriage over to the board, visually line up the part, and press the part down into the paste, causing the suction to stop and let go of the part. It’s still very much a manual process, but having the parts all ready to go in the bins and having several boards lined up in the brackets makes parallel assembly go a lot faster.
For large quantities of boards, there are automatic pick-and-place machines. Most electronic components come packaged as plastic or paper tape: each component is in its own little “cell” in the tape, with a plastic cover holding it there. The tape can be fed into a device that feeds the tape one component at a time, peeling back the plastic cover to expose it for the machine to pick up automatically and place it in the right spot on the board. These machines can run pretty fast, and even use cameras to verify that the parts have been placed correctly on the board. An automatic pick-and-place machine doesn’t save much time with small production runs, though. First, in order to feed components into the machine, the tape needs to be on reels. That’s not a problem if you’re building a lot of boards, because you’re using thousands of each component. But if you have, say, 25 of a particular chip, that’s just a short strip of tape. You could get that on a reel, but distributors usually charge extra to do that and it’s a lot of extra packaging (the reel and leaders on the tape are a whole lot larger than the little strip that actually has components in it.) Second, the automatic pick-and-place machine has to be programmed to tell it where all of the components go, what each component looks like (for visual confirmation of placement) and so on.
After the parts are placed, the solder paste is melted in a process called “reflowing.” There’s nothing fancy about this; it just involves heating the board up to about 250˚C for a short amount of time to melt all of the solder. There are fancy machines for doing this that carefully control the temperature profile, heating everything up at a predefined rate, staying at the peak temperature for a set time, and cooling the board off. If extremely high production yields aren’t necessary and one is willing to troubleshoot a few boards, the precise profile isn’t really necessary. People actually reflow boards with pretty good success rates using toaster ovens and electric skillets.
I assembled the first board by itself to verify that everything worked (it did, for the most part, except for a few small goofs) and then built a set of five more. The first board took a few hours to put together, and the set of five took a little bit more than 5 hours, so just about an hour per board. That’s not too bad. I’ll be able to finish up the twenty boards I need tomorrow.
Here are the six done so far. The big boxy thing at the upper left of the boards detects dust in the air. One of the small metal cans to the right of it detects ozone gas, and the other one detects a variety of nasty vapors. There’s also a high-resolution barometer, which can indicate altitude if the pressure at a reference height is known, and a light sensor that detects both intensity and color.
About those small goofs: when designing boards, one has to do a whole lot of work before anything is ever tested. The schematic has to be drawn based on the datasheets for the parts used. The footprints for those parts on the board need to be drawn as well, again from the information in the datasheets. The schematic needs to get turned into an actual layout of parts and conductive traces on the board, which means placing and routing things so that no more wires have to cross than there are layers in the board. Only once all of that design is done can the designs be sent off for manufacturing. There are a lot of places to make mistakes—there’s a lot of information that has to get taken from the datasheet and entered into the CAD software, leaving plenty of room to make small errors. Datasheets can also sometimes be misleading or wrong. On this particular board, I accidentally swapped the function of two pins on a chip, so the traces on the board should have been swapped around. The fix for something like this, without having new boards manufactured, is to cut the offending traces with an exacto knife and wire around them with fine-gauge wire.