Month: April 2024

This week I was mostly invested in experimenting with the mics because those are still not done and I think that’s more important than iterating on the casing.

After our presentation Monday, Prof Fedder met with us to look at the peak detector after the mic output because he thought our graphs looked different than what he expected. Turns out, it was the usual culprit of bad connection due to not soldering things because we didn’t want to make something permanent before we knew if it worked. (Ironically, we would still not know if things worked because the bad connection results in unstable results.) So after I soldered on some female header pins to go along with the male ones on the board, we got some much more stable results on the mic. This led to needing to double check the mic resistor values we used, as Katherine thought there wasn’t much difference between the 10k and 33k earlier, but now we can’t trust that because of the poor connection being a confounding variable.

She can probably tell you about those shenanigans more in her report, but the reason it matters to me is because the PCBs were expecting SMD resistors, so if the values in the PCB lab don’t match up with the ideal resistor values, then we were going to need to be creative in order to get that value. For example, the highest resistor value on the datasheet is 33k, but we only have 10k and 100k. In particular, I tried 2 approaches: stacking SMD resistors (three 100k resistors in parallel is about 33k) and soldering through-hole resistor legs to a SMD pad (can directly pick a 33k through-hole resistor). Theoretically, it’s also possible to do things like place the resistors in teepee style or Stonehenge style to be in series, but stacking them directly on top of each other was already a lot of work because of how small they were: they are easy to nudge out of alignment, and stick to the tweezers. Both the approaches I tried worked when tested with a multimeter! This is great because that means we don’t need to order new SMD resistors, which would take more time and delay us even more.

Funnily enough, we ended up not needing to stack resistors, because we discovered that the 100k seems to be more sensitive than the 33k in providing a larger range of values (which would be helpful for scaling/converting to dB). For 33k, the range was very small, around 50mV, but for the 100k we had about 300mV range.

So the next step was making the new PCBs. We are trying out the hard boards first to sanity check some things such as the new mic resistor value. This time, whoever made the stencil was clearly not doing their job properly, because the offsets between each of the 3 board strips were larger than the actual distance: basically, we could only align one strip at a time 🙁 This meant needing to pre-cut the boards before they went in the oven (otherwise, the stencil would smear previous solder paste when we tried to move on to soldering the next strip), which leads to the PCBs being smaller and later, which lead to the middle PCB flying off the rack and being a sad pile of unsoldered parts (the lights, capacitors, and one mic fell off before they could be soldered). Trying to spot-fix things with a heat gun didn’t work because the air would blast away all the other pieces out of alignment before the solder could melt. Next time, we will try to group all the PCBs nearer to each other so hopefully they will not be blown away. This may not be a problem with the flex PCB because the board strip can bend away from the stencil after solder paste is applied to avoid subsequent smearing, so we can keep the PCB in one piece in the oven. Another tidbit about the new boards: because of the Fusion360 grid snapping pain I mentioned a few weeks ago, the right mic of the board is a little bit closer to the LED pad than the left mic. Hypothetically, it should still be enough room to not interfere. Based on mic validation tests, the right mics of both boards didn’t respond to anything compared to the left mics. But we also only have 2 boards so far, so it could be due to small sample bias.

Unfortunately, even for the left mics, when we tried to recreate the large range we saw earlier with the new PCB it went back to being around 50mV, which is very sad. Not sure how much of this is due to the different testing environments, since the first room had a lot of background noise and the second room didn’t.

In terms of scheduling, the stuff I’m officially tasked with, which is casing construction and Arduino code integration, are still blocked from being finalized by the whole “things aren’t all working yet” that I’ve also been talking about for the past few weeks. I made a basic mock-up of the case already, but I can’t try sealing it yet because again we are still using the boards for testing, so it’s more convenient for the board not to be in the case for that part. So in reality I’ve just been continuing to help with trying to get the mics to work. For the next week, if the mic (and battery, beetle, bluetooth…etc) issues are resolved, then I will be making a lot more PCBs, presumably both flex and hard PCB just to be safe. And then I will finalize the casing structure with the board inside.

The two main things I’ve done since the last progress report are related to the PCB, case assembly, and testing.

First, I remembered that I have IDEATE access so I went to the physical computing lab to look at what they had. Happily, they have the fast-switching Zener diodes that we need, so we don’t need to order those anymore. They also have a variety of op-amps (which we thought were needed at the time) and Polulu boost converters so we don’t need to order any of those anymore, and I also took some buttons and switches. Then I also went to the PCB lab to count out the SMD resistors and capacitors to make sure we have enough of all our values. I labeled each strip since there’s no markings on the components themselves.

As a side quest, I also noticed that some of the header pin connections weren’t adequate (since the holes were still visible), so that made me wonder if that’s the reason the lights don’t work for all the boards. Especially since the lights are big enough and the soldering pads are far enough away that I can clearly see that SMD solder connections shouldn’t be the issue. After touching up the connections, we now have all the board’s lights working properly! Woohoo! That means the success of the boards only depends on mics, so by redoing the calculations, we’d need about 10 boards to be safe. Turns out the minimum number we can order is 15 so this will be fine (we bought enough Neopixels for about 28 boards so this will be fine ish, even with 2 designs which would be 30 boards total). I also took this opportunity to demonstrate how to solder to Katherine so she could try different resistor values for the mics (even electrical tape had a slightly sus connection, so soldering was necessary to properly test).

Next was redesigning the PCB to include the peak detector. For a while, we were debating whether we needed the op-amp. I preferred having as few components as possible since we were running low on space and it was going to be difficult to route things around such a large object. Prof Fedder also drew a simpler diagram where we only needed the “top half” of the peak detector, since we didn’t really care about enveloping the bottom-half of the signal, which further reduced the number of components in the original diagram by half. After Katherine did some simulations in Lspice, she confirmed that we don’t need the op-amp, thank goodness! That made my life easier. What didn’t make my life easier though, was Fusion 360 being weirdly different since the last time I designed it. For some reason, some components just didn’t snap to the same place in the grid as they did in the previous design, even though the grid granularity and components were the same. This was solved somewhat by rearranging the component order and then putting them back, or just sucking it up and saying “this slight difference shouldn’t affect the functionality” and moving on. Katherine did some calculations for the trace width resistance and ordered PCBs.

Cutting the plastic for the casing was also such a big pain! The tubing was decently easy to slice along the open edge, but the knife (and all other pointy objects I could find) were all not sharp/strong enough to puncture the tubing for cutting holes. I tried drilling to make the hole bigger, but the drill bit ends aren’t sharp enough to do much 🙁but it made the hole slightly larger so I could try fitting a small wire cutter in there to widen the hole. It took so long, and the hole ended up looking very jagged. Later, I realized that the battery couldn’t fit through the opening, even when squishing the pipe, and especially not if you need to add a PCB on top of that. So we’re not going to be using the tubing anymore.

The back-up option was the rectangular plastic wrap, which we’d need to measure and cut and glue together again. This is more work than being given a tube, but it gives more customization options (and the plastic is also much thinner). It has the same problem as the tubing where it’s pretty puncture-proof (this is pretty good for durability, but it’s a pain when we’re trying to make holes), but a helpful Techspark person showed how to use the hole-puncher, so we were able to make some holes where the mics would be. I also tried the method of just gluing the edges of the top and bottom pieces of the plastic together, and testing the durability from bending. It starts off ok, but it takes a lot of glue to seal it well, and over time the glue cracks and the edges start opening.

So for the next iteration I tried overlapping more surface area (thus, making the top piece of the casing much larger than the bottom piece). This time, I calculated how much I’d need for each piece instead of guesstimating. It’s hard to guess exactly how much I need because there are some variables that are hard to measure, such as the curve of the top piece. However, I think the sizing is all right because the height of the bracelet can be adjusted based on how much is overlapped with the bottom piece.

Prof Fedder also suggested that we need some netting to help protect the mics, so I found this plastic netting that we could use.

Originally, we planned to be done integrating (including the case) and onto testing by now, but of course that hasn’t happened. It’s hard to know if the casing design I have now will be greatly affected by having a flexible board instead, but I’ve discovered so many roadbumps in creating the casing that I think it’s important to keep going through the building process, even if it needs to be with a hard PCB. I aim to have a case finished by Wednesday (as in, the case is glued together, has the correct number of holes, and fits the battery and PCB with some room to spare). If the flex PCB arrives early enough, perhaps I will have time to test the fit and make adjustments before the next status report as well. I also plan to assemble the PCB with Lucy like last time, and also try to debug the button press.

Throughout this process, the way I’ve learned new things is through a few methods. Mainly, I try searching for tutorials (ex: on YouTube for PCB design like creating library components), and reading forums for how people have solved similar problems). In addition, I think an important method is knowing who’s an expert and consulting them for questions. The obvious one is our Prof and TA, but this also includes classmates (who are working on similar projects), and Techspark people (since they know what specialized tools are available and how to use them).