The goal of this project was to replicate the popular color organs of the 1970s using analog electronics only - no microcontrollers or digital logic. Once prototyping was completed, I designed and ordered a PCB for the project. See the project report here and a video of it in action below.
This project is currently installed in the bathroom of my dorm, and can synchronize to the hall-wide audio system or display ambient effects.
Using an Arduino with a bunch of relays, I built a room automation system capable of remotely controlling the lights, projector screen, and door lock. The system also has a webcam to remotely monitor the room. Boxes with 4 outlets each control miscellaneous devices. The website was built following tutorials from W3Schools, and is also the basis for this website.
The goal of this project was to project image and text notifications on a desk and interact with that projection via computer vision. The physical hardware platform for the project was a robotic arm, disguised as a common desk lamp, with a camera and projector at the end, all controlled by an FPGA. Notifications come from a connected computer and are keystone-corrected when projected, to eliminate distortion caused by projecting at an angle.
I worked as part of a team of 3 on this project, being responsible for the keystone correction, serial communication, and sound effects. These portions are shown in green on the block diagram below. View the final report here.
I built this controller with the intention of having an installation of 10 strips of 3528 LEDs on the ceiling of my room. After 3 of the 4 lights in our dorm bathroom stopped working, however, I decided to prioritize an installation in there instead. The sticky back of the LED tape fit perfectly between the tiles, and it is now affectionately called the RGBathroom. This was the first board used in it, and could only run through ambient color effects, unlike the Analog Color Organ project which replaced it.
I, as part of a team of 4, designed, built, documented, and maintained a powered joystick arm for people with multiple sclerosis. It allows the controller to be retracted electronically, without requiring physical strength to push it out of the way. The device is still in use at The Boston Home today.
This project examines ring resonators and how to efficiently build one from laser cut acrylic to operate in the UHF band. The final system provides 15 dB gain at the resonant frequency with the ring compared to without the ring.
Read the project report HERE.
This laptop was found in MIT's e-waste recycling. Nothing appeared broken, but when plugged in it would try to draw more current than the power supply could source. After taking it apart and verifying that the connector was not shorted, I started probing components on the motherboard. Eventually, I found a capacitor that was shorted, likely by the metal plate below the keyboard which can be pressed down to touch the motherboard. Replacing it and reassembling fixed the laptop.
In fall 2018 I had a class 1.5 miles away from the class immediately after. This dilemma lead me to want an electric scooter, as I could easily carry one into class but also move faster than possible running. I initially bought a broken Razor e300 scooter for $20 to fix, but got lucky and found this one abandoned and covered in rust in the recycling. After finding suitable batteries, removing rust, replacing switches, and attaching a light, I had a working scooter! Charging the batteries I have requires opening the frame, so I designed and ordered a PCB to facilitate the charging process with the charger I have as well as provide headers to monitor the battery voltage as I ride.
In my dorm we get a large amount of freedom in what we do to our rooms. With access to MIT's recycling, I was able to put some unusual but nice features in my room. Large cabinets removed as part of a building renovation hang from the walls, a desk spans the furthest wall of the room without any legs, and a projector screen hangs in front of the couch.
There is also laminate hardwood flooring in the room now, hiding the ugly yellow tiles underneath it.
At the beginning of each year, my dorm builds a bunch of wooden contraptions to attract freshmen interested in our maker culture. In May 2018, I was the construction lead for one of these projects, a large and very strong catapult. This meant updating the CAD from 2015 when the catapult was last built as well as directing the construction of the device. The whole catapult was built in a week of August 2018 and worked exceptionally well.
After leaving my dorm and moving to an apartment I wanted to set up a similar automation system. Without ethernet between each area of the apartment, the system needed to be wireless. I did this using the cheapest components possible, a system called SmartLife.
The software was awful and unable to reset triggers automatically when detecting motion. Rather than start over with a more expensive system, I reverse engineered and modified the sensor itself to accomplish this functionality.
I believe the sensor used is a PIR612, but with no part number I cannot be sure. The board differs substantially from the typical PIR Arduino sensors. Luckily, I was able to find a datasheet with a similar application circuit and modify its delay and sensitivity appropriately for my application.
For a controls class midterm project my team built an asynchronous sensorless brushless DC motor controller. The goal would be to attach it to one of our homemade scooters for demo if it worked well enough. We reused 3 extra boards of a buck converter PCB I had designed and manufactured for a previous class to save time since grading would be based on controls.
This was probably a mistake. For the controls to work well, a clean measurement of the phase difference between voltage and current is critical. With the lengthy wire runs from the bodged hardware, however, the entire system was incredibly noisy and could only spin a small quad-copter motor.
(I should also mention that by "team" I mean myself and a very good friend who also builds lots of stuff. We realized early on it wouldn't work well but with 1 week allotted for the entire project we didn't care to make it better)
Purchasing a used car meant the stereo was outdated, lacking bluetooth support or voice controls. Being an Audi, however, the sound system was designed by Bose and entirely nonstandard. I reverse engineered the system to find a way to upgrade the radio to a standard one without sacrificing sound quality.
The amplifier is not built into the head unit, but instead into the base of the car. Further, the alternator output is not filtered so the input is super noisy so any high impedance inputs or outputs will be noisy. Insted, the amplifier expects a very low impedance output while still being a lower voltage. This meant that the line level outputs on the stereo I purchased couldn't be used without sacrificing quality. The solution arrived at was twofold:
1. Filter the input power to isolate noise into the head unit
2. Use a line level converter transformer to convert the low impedance speaker outputs to low impedance line level outputs
Amplifying the signal twice isn't ideal, but the sound quality is better than that from the native line level outputs where you can hear interference from the GPU and touchscreen activity.
At the beginning of each year, my dorm builds a bunch of wooden contraptions to attract freshmen interested in our maker culture. In May 2018, I was the construction lead for one of these projects, a large and very strong catapult. This meant updating the CAD from 2015 when the catapult was last built as well as directing the construction of the device. The whole catapult was built in a week of August 2018 and worked exceptionally well.
At the beginning of each year, my dorm builds a bunch of wooden contraptions to attract freshmen interested in our maker culture. In May 2018, I was the construction lead for one of these projects, a large and very strong catapult. This meant updating the CAD from 2015 when the catapult was last built as well as directing the construction of the device. The whole catapult was built in a week of August 2018 and worked exceptionally well.
A python webpage to display Dallas DART light rail real time arrivals. Real time data is not available through DART GTFS feeds so it is scraped from the mobile website instead.
The code can easily be adapted for any DART rail station. Finding a train direction relies on knowing the ending destination, so the dart website link as well as destination stations will need to be updated. For a station as far north as Spring Valley this is simple, since there are no turn-back points between it and Parker Road. For other stations with multiple headsigns in both directions, all possible and turn-back points will need to be added to the northbound/southbound if statement.
The code is available on Github here. I also have a precompiled image for the Raspberry Pi Zero W since Google font + firefox-ESR setup became annoying on Raspbian Buster.
This slide from October 2020 leads me to believe feeds aren't going to be published anytime soon, but it is encouraging to know that they are considering changing the policy.
For my thesis mid-pandemic I needed access to a laser cutter. The easiest way to do this was to purchase the cheapest model of CO2 laser cutter available and modify it to fit my use case.
The $280 K40 laser cutter fits the bill, but its primary purpose is to make japanese stamps. In order to use its "40W" CO2 laser on other materials it needed:
There's so much more to say about this than makes sense to fit on a page like this, but I don't want to get too deep into it as all of the interesting quirks of this machine are well documented by the community and summarized at K40laser.se