Although these boards are #3 and #4 in the number of components, with 425 on the ID board and 368 on the ALU board, these lack the repetitive DRAM grid pattern of the previous boards. Since I do the parts placement myself, I made sure that all the transistors are oriented in one of two directions, rather than the four possible. This minimizes the need to rotate the board repeatedly as I remove the parts from their tape carriers with tweezers and carefully place them on the board.
This also helps avoid getting the orientations wrong. While the orientation of the three-lead FDV301s is pretty obvious from the silkscreen markings, the four-lead BSS83s is not so obvious. I'd already populated the OPR and OPA register circuits when I realized I'd placed all 14 of the BSS83s in those circuits rotated 180 degrees. I fixed 12 of these before reflowing the solder paste, but during a later visual inspection found I'd missed two which had to be unsoldered, rotated, and resoldered by hand.
 |
| The Instruction Decoder board |
When assembling the Instruction Pointer and Scratchpad register boards I'd tried splitting the process so I could reflow the solder on the right half of the board before starting to place components on the other half. This had sort-of worked for the IP board but had been a mess on the SP board. This time I decided to stencil out solder paste over the entire board first, then place all the components in one long marathon session. This appears to have been the right decision, even though it took four hours of continuous work to complete.
The only part I'm really not happy with was the reflow process. I used a household heat gun to preheat the board from the bottom until the top reached about 100℃, then used my hot-air rework wand on the top to do the final heating. The result was uneven heating, inconsistent solder reflow, and a slightly warped (twisted) board. I then had to go back over the board and touch up far too many joints.
I suspect a lot of this inconsistency is my use of via-in-pad for ground connections. Just about every transistor has a pad with an embedded via connecting to both of the ground planes, which form large heat sinks. Frequently these were the pads were either incompletely wetted or not reflowed at all.
Maybe my heat gun can be cranked up high enough to melt solder, but I'm afraid the airflow it generates would blow components off the board. A good alternative I've been eyeing would be a homebuilt reflow oven made from a convection toaster oven, but I don't make boards often enough to justify that kind of effort.
While websurfing last night I came across a Sparkfun article from 2006 entitled Reflow Skillet. They describe using an electric skillet to reflow PCBs. This avoids the problem of air blowing components off the board as the board is heated from the bottom by conduction. This would fix my problem of heat being wicked away by the copper ground and power planes, as these will be at the same or higher temperature as the top layer.
The key to seems to be finding a skillet that will heat fast enough to a temperature that the high enough to melt solder before the board and components are damaged. The solder I'm using, ChipQuik TS391AX50 (63% tin, 37% lead), melts at 183℃ (361℉). Most electric skillets I saw for sale maxed out at 400℉ (204℃), and folks who have attempted to use these reported disappointing results. So I was careful to look for one whose temperature control goes to 11 ...err... 450℉ (232℃) like the one used by Sparkfun.
I found one on sale at WallyWorld that seems to meet these requirements. It's a refurbished unit and cost me all of $28 including shipping. It's supposed to show up tomorrow. I'll probably try a few experiments with one of the extra Instruction Decoder boards and a few FDV301 transistors scattered about to see how well it works before risking a fully-populated ALU board with 110 irreplaceable BSS83s on it.
No comments:
Post a Comment