High quality sound should be a right, not a luxury. To that end, I worked to make a cheaper, open-source version of cobaltmute's pupDAC for a EE PCB design class.

Over the course of a semester, we selected a circuit, analyzed it, turned it into a MultiSim schematic, turned that schematic into a board layout , and then ordered and built the board. I knew that it would be a challenge to build this board, as I was targeting a very small footprint (2"x3" for 80 components) and a relatively complicated design, especially for my first board layout ever.

Soldering Challenges

The following video shows a timelapse of me constructing the board, which required great soldering dexterity.

For my board, I started with a solder paste mask, and spread solder paste over the top surface (0:01). Then I used tweezers to place the parts, and stuck it into a reflow oven. Here, I discovered that our school's default temperature profile was inadequate for my board, because it essentially flash-flowed the oven. It didn't have a soak time, which meant that the ground plane didn't have enough time to heat up, and any parts connected to the ground plane (like all my capacitors) needed manual soldering to melt the paste.

Even with this complication, It would've been a simple job, except that I ran into a nasty problem after I had already started assembling the board: the main IC, a USB to I2S audio data converter, was too large for the footprint (I later traced the error to MultiSim having a too-small footprint for the TQFP-32 package).

So, like any enterprising student whose grade relies on his circuit working, I decided to take the logical next step: solder jumper wires directly from the pads to the leads of the IC (no, I couldn't have ordered a new board, I only had a week till the semester ended, and I still had to redesign the board).

Even though both my professor and our lab technician recommended against this jumper idea, I decided to carry on with my audacious plan. I initially thought I could find some 8-wide ribbon cable, and solder that to the pads, and then fan that out to the IC itself. However, the insulation was too wide, and I couldn't get the leads to stay well-joined to the pads. At this point, I should probably note that the leads of a TQFP-32 are spaced .8mm apart, and the pad that I was soldering onto was spaced at .5mm apart. The next step was to get some wire wrap from the school's technician, which was between 30-32 gauge, and JUST BARELY fit onto the pads (I believe they are 10 mil pads).

I couldn't find wire strippers that worked for wire that small, and using my teeth was ineffective (I ended up bending the wire more than stripping it). So I figured out a way of stripping the wire that was more effective than any other method I found yet, with a yield of maybe 50%... I used a knife to cut the wire to length (roughly 1/4"). Then I would roll the wire under the blade while applying very light pressure. Half of the time, this would create enough of an edge for me to pull the insulation off with a fingernail (the other half it would just cut straight through the wire, and I'd start over). This can be seen at 1:08 in the video.

After making the 32 jumpers needed to connect the IC to the pads (and a few extras for the inevitable problems. And then a few more...), I got down to the slow and touchy process of actually soldering them to the pads (0:47). After getting them onto the pads (1:17, a process that wasn't totally included in the video), I then had to somehow carefully bend the fragile fragile wires up (without breaking them or tearing them off the board), and then solder the IC onto them without heating them too much that the pad end of the wire would come loose. Which happened more times than I would care to admit...

Once I found a rhythm, it was fairly easy to get going, but I still ran into some problems. One was the above-mentioned wires coming loose. If the wire was on the end of a set of 8, it was no problem. I'd just put a new one on. But if one in the middle came loose, I had two choices: desolder all jumpers between it and an edge, put a new one in, and then resolder all of them. Or I could run a jumper directly from the next point in the circuit to the IC (such as a via to the IC). I used both solutions, depending on the situation.

I also found out firsthand that sometimes PCB traces/pads are able to become deattached from a board after several cycles of heating-cooling. Luckily, most of the pads that came up were either power or ground, so I was able to make this hack even hackier by joining several ground jumpers together, or just bridging a few connections.

Compared to the giant pain that the single IC was, the back was extremely simple. Because I didn't want to risk components coming loose in a second trip through the reflow oven (especially my delicate IC soldering job), I used a vise to hold the PCB steady, and then used a paste mask to lay down solder paste on the back side. From there, I simply placed the parts, and then heated them up. In the future, it would probably be a good idea to melt the solder as I go to avoid problems with bumping parts that have already been placed. The back side of the board can be seen starting at 1:54.

After completing these three big tasks (top side [reflowed], USB DAC IC [hand-soldered, jumpered up 1/2" in the air], and back side [hand soldered with paste mask]), the only thing left to do was test it. Unfortunately, it failed most of my tests. It was able to turn on, be plugged in, recognized as a USB Audio device in Windows, etc. But I only got clicks out of one channel, and no high-fidelity audio goodness like I hoped. I hoped to spend more time being able to test it, but unfortunately I ran out of time.

Testing

As previously stated, the circuit was able to turn on and enumerate on USB. Most of the voltage signals seemed to be correct based on the testing that I had time for. This is when I was really thankful for the 20+ test points that I built into the board. However, I was really frustrated with our school's Tektronix oscilloscope probes, as I was unable to reliably hook them onto the SMD pad test points that I was forced to use in the middle of my circuit, which would've helped the testing process considerably.

However, I was only getting intermittent clicks out of one audio channel, and was unable to diagnose the problem with an oscilloscope (and a little bit of time with a logic analyzer). My current hypothesis is that the circuit was correct, but the IC was damaged due to all the heat that it was exposed to in the drawn-out soldering process. Another possibility is that the audio signals weren't synced correctly, because the trace lengths were longer than necessary.

I hope to be able to redesign this board sometime in the coming months, order all new parts and a new PCB, and take another shot at soldering it all together. Especially knowing what I know about board layout and the complexities of this circuit, I think I have a decent chance at making it all work.

Time Estimates

The soldering process took about 15-20 hours over the course of a week (right before finals), and then the testing (what little I was able to do) took place over a few hours in the middle of finals.

Shout out to my school's microscope to check for bridges, and our lab tech who put up with me for many hours and even stayed late a few times to deal with this kid who refused to give up on a project that he probably should have.