Hi Everybody,

As Steve pointed out to me yesterday, it's been quite awhile since I posted anything to the boards. The truth is that my involvement with µ-controllers and/or embedded systems is limited to the extent required by my own R&D. Since I lack expertise in the field, especially compared to others like Erik, or Steve, or Craig, et al, I don't often have much to contribute. Still, I am glad this site exists when I need help.

That reminds me. Thank you Erik. Yesterday when I was puzzling through the simulation/debugging of my code, it was one of your posts on the Atmel forum that first alerted me to the issue of SBUF being two separate registers.

Anyway, I know that questions concerning prototyping PCBs often arise on this forum, so I thought I would share the following. It may prove informative for someone who is curious about prototyping a PCB.

This is the most recent version of the control circuit I've developed for my (still ongoing) project, a mass spectrometer.



I mill the copper features with an LPKF C60. As you can see, it is capable of milling features much smaller than I could assemble with tweezers and a soldering iron. According to the manufacturer (and if my memory serves) it is capable of producing 4 mil lines and spaces, though I am not going to pay for the tools (bits) necessary to do so. The truth is that I have produced similar (to this board, not 4 mil) scale circuit features with photoresist and ferric chloride (the old GC chemistry). But the chemicals have gotten difficult to find, and they're hazardous to keep and dispose of. So I made the switch to the LPKF.

As an example of small features I had to deal with, look along the right-hand edge of the board, not far from the bottom, and find U2. That is a SiLabs CP2102 USB to UART Bridge. Now look at the next picture. That is a close-up of the lands for that chip. That's a dime in the picture. Or if you're unfamiliar with US coinage, the four white corners (the legend) mark off a 5mm x 5mm square.



You may note that the same corners are included in the copper, though the legend is misaligned just a bit. Those are for chip placement alignment. Placement of the chip is one of the principle obstacles I had to overcome. In order to reliably place a chip so small I had to get a fine-pitch placement table, with a vacuum pick-up tube and fine adjusting knobs to adjust the x, y and Θ of the board relative to the chip. Even then I had to install a microscope above the vacuum pick-up.

There has also been mention made, from time to time, of using a toaster oven for reflow. In point of fact I used to use a Ransco lab oven, which is essentially the same thing. However, I discovered that with smaller features it wasn't sufficient. A lab oven or a toaster oven that is just turned on and used to melt the solder paste works okay if your solder paste is in large enough bricks (blobs), but on very small pads you will get a lot of failed joints.

The reason lies in the chemistry of the reflow process. It is necessary to heat up the solder paste enough to activate the flux so it can do its work on the metal surfaces. If the flux doesn't clean the surfaces of all oxides the solder won't whet. Then the solder must be melted, whetting to the surfaces and evaporating the remaining flux.

But with small amounts of solder paste and a slow rising temperature, too often the flux will be activated and evaporated before the solder melts. When that happens new oxides form (as soon as the flux is spent) and the melting solder can't whet (bond).

Interestingly enough, I improved my reflow process by switching over to a toaster over, a Black and Decker Infrawave oven. However, it is controlled by a thermal profile controller to produce nice consistent results. Ironically, my lab oven wasn't capable of ramping up the temperature fast enough to prevent flux obliviation for such small quantities of solder paste. The Black and Decker Infrawave was the only oven I found that could.

If your board doesn't include anything smaller than TSOP scale packages (50 mil pitch parts), Then you can probably use any old toaster oven to reflow your solder and get results reliable enough to not need anything more elaborate. But if your trying to solder 25 mil pitch parts, or something even smaller like the CP2102, you'll need to get a thermal profile controller.

What about placement of the parts? Do you need a mechanically and/or optically assisted placement device? That depends on the pitch of your parts, not to mention your vision and the steadiness of your hands.

But if you're using anything smaller than SOIC pitch you will need some way to stencil the solder paste onto the lands. You're not likely going to do it with a pneumatic dispenser.

On the other hand, if you are not trying to assemble anything smaller than SOIC, you can use a technique that I call excess wicking. Basically, you just glob solder onto all the leads down the side of the chip. soldering everything to everything. Then you use a solder wick to suck it back up. It will leave enough solder behind to leave the leads soldered to their respective pads. I've had good luck doing this down to about 50 mil pitch. But once you get down to 25 mil pitch it's prohibitively, frustratingly hit-and-miss.

Finally, a word about photo-processes. I still use photosensitive processes for my solder mask and legend. And since most people don't have a dedicated milling maching like the LPKF, many are still using the old GC and Kepro chemistries. Regardless of the chemistry, if you're trying to produce fine pitch features in photo-resist, you will need a good UV light source.

You will want to ensure two things. First, you want to make certain that you get even and uniform exposure over the entire surface area. Basically, this means don't put the light bulbs too close to the exposed surface. On the other hand, don't move them too far away either. The farther away they are the longer you have to expose. The artwork is never perfectly opaque, which means that the shielded areas do get some exposure. But since the artwork is very nearly opaque, the exposure of the shielded areas is more a function of time than the slight variation in intensity a few inches would produce. The bottom line is this. The longer you have to expose, the less reliability you achieve.

But there's one more thing that will make even more difference on very fine scale features. Refraction. Your transparency has artwork printed on one side. If that artwork is on the side opposite the exposed surface, light will refract under the artwork and expose the fringes of the feature you're trying to establish. If that feature is very small, you can easily obliterate the entire feature.

The solution to this problem is to make certain that the artwork is printed on the same side of the transparency that will be in contact with the board. This may mean having to print some of your artwork in mirror image. Do so and you can easily produce 10 mil lines and spaces, or even smaller with less reliability.

In short, you can prototype boards yourself, and you can do so on a student's budget. But you will have limitations and you should understand them from the beginning.