I started getting into a bit of woodworking a few years ago, and while I don’t do it that often, it is something I enjoy from time to time. Part of the challenge I have is the relatively small area in which to work in my garage, combined with the fact the space is used for many other things. This results in a rather significant amount of dust accumulating rather quickly in many places where it does not belong. I built this project to help mitigate this issue, in addition to the other benefits quickly ensuring the air is clean and healthy (though of course I do wear a dust mask!)
All of the standard tools I have are hooked up to whatever dust collection system is available for them. This works OK in some cases, and quite poorly in others (I’m looking at you miter saw). I felt as though having an open air filtration system would help this, and in fact it does an excellent job of rapidly reducing the PM2.5 count in the open air.
I discovered the filters made by Wynn Environmental, which were highly recommended and decided upon a 34″ MERV 15 filter, combined with a rather robust 800 CFM 10-speed fan by AC Infinity. The fan had a separate wired speed controller, but I needed a way to replicate the signals for use in my device. Reading through the product description mentioned it was PWM-controlled and hooking it up to my scope revealed the necessary details such as frequency and voltages. After playing around with a ESP32 device to simulate these, I needed to go back and confirm a few other items so I just used my trusty Fluke. And then came the mistake: I forgot to remember that I had left my meter probes in current mode. Hooking it up shorted the output of the wired controller and killed the driver IC… whoops! Looks like now I only have the option of finishing this project to use this filter.
Luckily for me, the prototype I made with the ESP32 worked rather well. Next, I added the other necessary components and designed a circuit: a PM2.5 sensor, a temperature/humidity/gas sensor (to combine uses and eliminate an existing one in the garage), then a display and UI. The last part was a little tricky, but I remembered I had a collection of several of TM1638 devices sitting in a spare parts box. It had several 7-segment LCD displays and a row of several buttons. And since I wanted this to be somewhat automated, it needed to be hooked up to Home Assistant, my home automation system. Looking through some libraries, I see one already written compatible with ESPHome for a TM1638. Great, this looks like it would be a perfect fit! I’d have to write a custom component to integrate it, and I haven’t tried that before but it looks like a great learning opportunity.
A few hours of looking through documentation and C++ coding resulted in a working MVP. I still needed to breadboard out all the components to complete a full test, but thankfully it went mostly smoothly. I only hit an issue with the opamp I initially chose, as it wasn’t able to output a strong enough signal for the fan to detect. My plan was to power this device straight from the fan controller cable, as it supplied a +9V for this purpose. I swapped out the opamp for a rail-to-rail one, and this got it working perfectly.
Next, I spent quite some time learning the basics of KiCad, as I wanted to order a custom PCB for this. I tinkered around with Fritzing before, but found it very limiting and extremely frustrating to use, so now seemed like the perfect time to learn how to use a proper EDA tool. The schematic was relatively straightforward, but the PCB editor was less so; looking back now, I think most of that was just understanding how all the various aspects worked. Custom footprints and syncing changes back and forth between schematic and PCB took a little bit of practice to grok.
Then came the ordering, which thankfully was very simple as I had chosen PCBWay as the manufacturer (I couldn’t begin to count the number of hobbyists promoting them), and there was an integrated plugin which made 1-button uploading extremely simple. A few weeks later I had my first custom PCB’s sitting in my hands. A little bit of soldering later and voila!… oh wait, no it didn’t work. Hmm. I debugged things and realized I had made a mistake or two in routing the traces. A quick fix, another order and a few weeks later I had an updated and working design.
Now, how to physically build the device, and by that I mean an enclosure. Since there were quite a few important details regarding layout, spacing and component sizes I had to ensure this was handled carefully. A first attempt using a hobby project box quickly demonstrated this approach would not suffice. I have a 3D printer and heard of OpenSCAD before, and thought that might an option, so once again I dove into learning an entirely new tool. OpenSCAD was actually not too difficult, given my background. But to write code to design a 3D model with all of these specific requirements would be a big challenge. So I found a library called YAPP Box for generating custom boxes. A solution seemed within reach! But alas, there was a rather significant amount of time spent hacking the library to work around limitations.
I finally came up with something which looked feasible. A few hours later, I had a printed model in my hand. And after some quick assembly and despite mocking up 3D models of them, it became apparent that I failed to accurately account for the dimensions of the internal components. And there was no room left for the necessary wiring! Back again to revisit the model.
The second version fit everything, but after assembly and testing I encountered another issue: since things were so tight within the enclosure, the relatively small amount of heat dissipated by the ESP32 was enough to drastically skew the temperature readings. A few attempts at drilling ventilation holes proved to be somewhat, but not completely effective. Finally, I decided to just mount the temperature sensor on the outside of the unit and this is ultimately what I went with.
The OpenSCAD library I used also supported button models to be generated, but once again the version I made did not have proper clearance for them. I’ll eventually (maybe?) go back one day to revisit and design smaller buttons, but honestly this works well enough as it is. The white stains on the front cover were from my attempts to CA glue & mount the buttons on the front face, but before I realized they would not fit.
I didn’t expect this project to be trivial, as there were a lot of new skills and applications I needed to learn to bring it to completion. However, a few things were rather surprising to me after the fact:
- I did not expect the enclosure to the be most challenging aspect of the project
- Despite modelling the internal components, I still made a number of sizing errors when it came to ultimately assembling
- Thermal dissipation had a larger influence then I expected
These issues would have been easier, or even completely mitigated if I had chosen a larger enclosure to fit everything. But I wanted a fairly compact design that would fit cleanly on a post in my garage – I think the biggest takeaway was the difficulties in meeting this constraint.
Leave a Reply