Lots of people have figured out surface-mount construction techniques and tools that work on a very limited budget. Getting boards assembled by a professional assembly house can be expensive, and there can be rules that may not make sense for a small project. (I've used an ISO-9000 board house in the past, and if I don't bring them paperwork that looks a particular way, they get antsy.) So these techniques for soldering teeny fine-pitch parts are cool, especially with the deals one can get these days on small runs of prototype PCBs.
If you're going to do this, do NOT use your normal food toaster oven for
soldering. Get another toaster oven just for soldering, and do NOT prepare food
in it. It will fill up with lead and foul chemicals.
My first experiment was to put some solder paste on two pads, put a resistor on them, and use the toaster oven to melt the solder paste. I used too much paste but the results were very successful. I am quite impressed with the toaster oven approach. Toaster oven discussions on the web talk a lot about the profile of temperature over time, and how semiconductors should not be subjected to extreme heat for more than a few seconds. My impression is that you want to run the thing a little below water's boiling temperature for a few minutes, to dry out any residual water, and then very quickly ramp up the temperature to the oven's maximum, watching the solder joints with the magnifier through the glass door. When the solder appears to have melted everywhere (and then giving it maybe another few seconds for the solder joints you can't see), turn off the oven and open the door so that things will cool quickly.
Then I put together the hot air pencil: a desoldering iron $10 with steel wool ($3, Home Depot) wedged into the nozzle, and replace the squeeze bulb with a tube ($2, Walmart) leading to an aquarium pump ($6, Walmart) so that hot air comes out of the nozzle. The trick is to jam as much steel wool into the nozzle as you can, otherwise the breeze will be so strong that it blows the SMT parts around, and it won't get hot enough to melt the paste. The steel wool provides more surface area to transfer more heat to the flowing air.
If you want a hot air pencil that isn't a cost-saving kludge, the Weller WSTA6 is available for about $60. It uses butane. I looked at one and it wasn't obvious where you get butane refills, or how you apply them. Boston-area people can get one at You-Do-It Electronics.
The hot air pencil seems to take considerable practice to use. Also the solder paste seems to be a bit cantankerous; I've read of people using brushes with it. I'll pick up one or two small paintbrushes and see if that makes it any easier.
There's a learning curve here, so be ready to do lots of inexpensive
experiments. Get a roll
SMT resistors and some
boards to solder them onto, and waste them liberally while you develop
I did the first board in a crowded roomful of hobbyists. It was fun, and it was a pleasant reminder of younger days, but the hustle and bustle weren't too helpful for the level of focus needed for SMT construction. I found it much easier, quicker, and more productive to do my SMT construction at home. You want quiet, lots of light, elbow room, and good magnifiers and tweezers. I haven't needed a microscope yet but it might be smart to start looking for one.
For the second board, I used a small paintbrush (the smallest in a set from the stationery section at Walmart) to apply paste directly to pads. On the second board, I decided to use the toaster oven twice, the first time to solder the resistors and caps, and the second time to solder the ICs. My reasoning is that (1) the Rs and Cs can handle heat better than silicon, and (2) this way, when I'm positioning the ICs, I don't need to worry that I'll accidentally push the Rs and Cs out of alignment. I don't think it's actually good practice, and I would be evil to recommend it to anybody who thinks they can populate their board with a single reflow.
One element where practice helps (and unfortunately nothing else seems to) is in knowing how much solder paste is right. The first time with the stencil, I put on too much, and studying it under the magnifier I worried that it might be too much, and the result was shorts. The second time I probably overcompensated, leaving only one or two grains of solder on some pads, and I'm concerned that some pins won't connect at all. Especially on the 144-pin FPGA, I doubt I have the eyesight or steadiness of hand to solder them individually.
Practice is important, and as I wrote earlier, it's good to create an
environment where you can get a lot of practice for a small amount of money. The
paste, the toaster oven, and the paintbrush are cheap, as are rolls of some
0805 and SOT-23 parts. It would be nice to find large cheap volumes of big
fine-pitch ICs. The cheapest I've found is this CPLD
for only $3.91, of $3.31 in lots of 25. It's in a 100-pin TQFP package with pin
spacing of 0.5 mm. Somebody should make up a board especially for SMT
construction practice, with no need for vias or a circuit that makes any sense.
I might do that myself.