imst

So I built a smart speaker which is kind of designed to directly compete with commercial products from companies like Sonos or Denon. The speaker connects to Wi-Fi and can stream music directly from Spotify and also a player can be connected to it through Bluetooth for quick easy offline playback. I CAD modelled every part for it in OnShape (viewable on GitHub) and then 3D printed it. I even created an acoustically dead enclosure for it using plaster of paris and PVA glue with which I filled up the sides of the subwoofer enclosure. The whole speaker is then wrapped around a wooden shield to make it look really nice and minimalistic and also save tons of plastic and printing time. It even features an app which the user can use to control it and change its settings, even it’s equalizer. It also features a dynamic EQ to account for the Fletcher-Munson curve (so it dynamically boosts bass on lower volumes), which a lot of high end commercial speakers do. All logic is programmed in python and it’s running on a Raspberry Pi inside. It features a Hi-Fi DAC+ to provide high quality analog audio to a TPA3255 amplifier which powers a dayton subwoofer and 2 mid speakers and 2 tweeters. The hardest part probably the volume/EQ managing part, which I figured out by implementing various protection precautions in it to hopefully protect the speakers in the long run. Making the wooden enclosure was also pretty difficult because it just took really long and took a lot of force, I also had to be really careful to not break it while bending it. I ended up tackling that pretty well though. I’m really happy with how it turned out, it actually sounds really amazing. I actually do use this project on a daily basis, which I’m really happy about, because it’s actually a project with a good use case! So yeah, I think that’s it. Overall it’s an opensource Hi-Fi quality speaker that anyone can build!

This project is a mobile phone app for android made for connection with my Nexo smart speaker (other project). The app is built in React Native in JavaScript. It uses multiple screens. A part of the home screen is a wifi scanning function which scans the wifi for nexo speakers. It then saves the found speakers and caches them so next time the app opens its loaded quickly. You then can access each speaker in the menu and control it independently and change each of its functions, such as volume, equalizer, network connection, etc. The app can be toggled between light and dark mode and also has an english and a czech version. On the home screen you can view realtime playback of the speaker, which song is playing. When spotify is playing, it even shows the art cover, for bluetooth music it doesn’t. It shows the songs metadata and the speaker which it’s playing on. It’s made to support a master-slave speaker configuration for multi-room function, where on the home screen you’ll be able to only adjust the volume of the whole system and then in each speaker setting you’ll be able to control them independently. I had issues that I had to solve when making the network scanning function, which I ended up fixing and making it work. At first I had issues with it even running at all, then it worked fine in an android emulator, however when I built the apk, it didn’t work again. After using promises with the web requests and after giving specific network access for the app using app.json in Expo I finally got it to work and now it works great even when built. It’s now more rigorous and scans multiple subnets where the speaker could be hidden. Overall this app helped me learn a lot about app development and also about the specific permissions the app needs. It’s very cool to see the app finally work in the end.

So I built a testing Django café management website which helped me deepen my understanding of Django. The hardest part about this was probably figuring out the whole architecture and thinking about how to make everything work together. Which I tackled by creating two seperate web pages, one for staff, the other for customers, which are independent from eachother. I like how it turned out, because it looks simple and minimalistic while still being functional and having everything work. Another pretty difficult part of this was also making the qr codes work on the staff page, because after django loads the html, django isnt initialized anymore so i can’t dynamically create the qr codes like that, so I solved it by generating qr codes for all tables into a hidden part of the webpage and then fetching the qr code images from that into the actual order cards. This could’ve been done differently though. I also had to write a decent amount of javascript code on the template html files to make it work fluently with the backend and make it fetch in realtime and update every few seconds. I like how it turned out, I also learned a lot of things by doing this for actual website production. Because I wanted to dockerize it and make it docker ready I also learned a lot about docker and now I finally understand how it fully works and why people use it very often.

So I built a smart speaker that is able to connect to Spotify and directly stream songs from it using the Spotifyd library (which uses librespot, thanks a lot to librespot, because Spotify seemingly does not care about small developers) and also it can work as a basic bluetooth speaker to which you can connect to using any bluetooth transmit device using the aptX codec! The whole smart speaker uses an RPi 5 as the brain (could even be a zero 2 w probably) and it also uses some LEDs as feedback for the user (mainly about volume, but also skipping tracks, pausing/playing, restarting spotifyd (in case it breaks), kicking a user from bluetooth or force switching to bluetooth mode while Spotify is playing. All of these controls are controlled using 3 key switch sensing modules (basically buttons that can sense fingers a few mm away from it so it can sense through wood). The software part uses Linux for the audio processing, where I’m using pulseaudio with some optimized settings for high quality resampling (44.1 kHz to 48 kHz) and I’m also using Carla as a type of DSP to edit up, equalize (and also add more bass, because I love bass) and split the sound accordingly (so that 20-200 Hz goes to the subwoofer, 200+ to the midbass speakers and 2000-20000 to the tweeters) so that it sounds really nice. I also spent a lot of time making the plastic enclosures, gluing everything together, figuring out and fixing some structural issues I’ve encountered, and mainly, what I’m really proud of, is making the subwoofer enclosure. I filled the sides of it out (they were not filled, but had air in them) with a plaster of paris mixture with PVA glue to make a really nice acousticly dead heavy enclosure, so the subwoofer has an optmized enclosure to produce bass notes in, thanks to this, I managed to get the low tuning frequency down to 50 Hz, which is lower than the subwoofer is rated for! Overall this was a really cool project and I actually think I managed to make a speaker that sounds better than most smart assistants and actually is able to fight with even more expensive speakers that Sonos offers. The main challenging parts were figuring out the audio software things, because I’ve had a lot of problems while doing this project with Linux and audio issues, however I ended up figuring them out and making it even better than what I had at the start. I also had some issues with hardware and gluing things together and making the subwoofer enclosure airtight, but I figured out those things too, somewhere I had to make and print a new part, or add a bit of silicone. Overall I’m really proud of how the project turned out and I’m really happy with how it sounds. A lot more documentation and information is on my GitHub readme, so please check that out too!