Alex Bustos

Machined Wooden Pencil Case

Wed, 16 Sep 2026 00:00:00 GMT

Has this ever happened to you?

Recently, I tossed my rOtring 600 mechanical pencil into my bag and headed to a local coffee shop. All was well, until I got to the coffee shop. I opened my bag to retrieve my prized writing utensil, and was greeted instead with disaster: the metal tube at the tip of my pencil was bent.

This actually did happen, and actually is part of the motivation for this project. Around the same time though, I’d joined a local makerspace and was searching for a project to get familiar with some of the tools there. So, I figured I’d kill two birds with one stone, and try to make a case to protect my pens and pencils from situations like this.

Design

The design for this project was straightforward: a two-part case with two slots for rOtring pens or pencils, magnets to hold the case together, and a lip to align the two parts.

I iterated on the details a bit via 3D printing and made some small refinements here and there, but I didn’t significantly change the general shape of the design from the original concept. The final design is below.

Manufacturing

Pencil case being milled.

Manufacturing was fairly clear-cut. I machined both halves of the case from a single piece of maple, using blue tape and super glue for workholding. The biggest challenge I ran into was just working around the age of the Techno LC 4896 CNC router at the makerspace.

This machine is probably in the ballpark of 20 years old, which presented some interesting quirks. Most notably, 2D and 3D adaptive toolpaths don’t work, presumably due to the specs of the control processor. I ended up just using 2D pockets for everything.

I also ran into a problem that I would categorize as probably more of a skill issue than anything else. The first time I tried milling out the case, I wound up with this huge gap between the two parts, which are designed to be flush.

Gap between case parts.

The culprit wound up being a discrepancy between setups in my CAM. This machine doesn’t have a tool setter, so the way to set different tool heights is to touch off the tool on the top of the stock. I had two setups for this project: facing and everything else. For my second setup, I’d set my stock mode in Fusion to From Preceding Setup with Continue Rest Machining checked. I’d interpreted this as “the stock for the current setup is whatever’s left from the previous setup”.

This turned out to be true for toolpaths, but importantly, these settings do not update the stock box points, which still come from the original stock definition. And because I used one of these points as my origin, whatever height got faced off in my initial setup wasn’t accounted for in later operations.

Fortunately, the fix for this was easy— I just set the stock dimensions for the second setup manually. With that out of the way, the rest of the machining went smoothly.

Chamfer operation using 3D printed workholding.

To machine the chamfers on the exterior of each part, I used a 3D printed jig for workholding, shown above. The milled parts prior to sanding and finishing are shown below.

Machined, unfinished pencil case.

Wood Finishing is Hard, Apparently

After machining the case, I sanded the parts using progressively finer sandpaper, up to 320 grit. At this point, I felt like I was almost done with the project; in hindsight, I realize I was getting ahead of myself.

I chose Danish oil to finish the case, partly because I’d read it was the most foolproof option, and partly because an oil-based finish seemed like it would be easier to apply than a surface finish like wipe-on poly to parts like this with small channels and crevices.

The problems I experienced with the finishing process were almost entirely due to insufficient research. It didn’t occur to me until halfway through the finishing process— after I’d already applied several coats of oil— to watch a video demonstrating proper application. Instead, I read a few written explanations, one of which used the phrase “flood the wood”, which for some reason really stuck with me. So, for the first few coats I did exactly that: I poured copious amounts of finish onto the parts, trying my best to flood them with oil. Some of the aftermath is visible below.

Finishing setup.

This mistake seemed to be more messy and annoying than anything else. After waiting about 15 minutes for the wood to absorb the oil, I would still wipe off any excess, so— as far as I can tell— these early coats didn’t ruin my parts; they just made for an unpleasant and tedious application process.

Eventually I watched a video, figured things out, and applied later coats by dipping a clean shop rag into the oil and then wiping it onto each part, which made for a much cleaner experience.

Between the last few coats, I also used 400 grit sandpaper and #0000 steel wool on the outside of the case, which gave the exterior a great finish. Unfortunately, because of the geometric complexity of the inside of the case, I wasn’t able to give it the same treatment, so the interior is noticeably rougher than the exterior. This, in my opinion, is the case’s biggest shortcoming. It’s hard to notice at a distance, but can certainly be felt, and can make the case a bit hard to open at times. I’m definitely going to be keeping this in mind for future projects.

Between testing out finishes, making mistakes, waiting for oil to cure in between coats, and just generally dragging my feet, this process took me several weeks to get to a finished product. Finally, I used CA glue to fasten magnets into each case part. And with that, I called it done.

Conclusion

This was a fun project. Although it’s far from perfect, I use this case daily to carry my pen and pencil. It works well enough, and I’m happy with what I learned.

Finished pencil case.

The Mountain in the Sea

Sat, 05 Sep 2026 00:00:00 GMT

Premise

In The Mountain in the Sea, marine biologist Dr. Ha Nguyen works with humanity’s only (allegedly) sentient android Evrim and Mongolian mercenary drone operator Altantsetseg to investigate what they believe to be a civilization of intelligent octopuses near the Con Dao archipelago off the coast of Vietnam.

Some chapters focus on two other characters whose storylines loosely intertwine with the main plot: Rustem, a Tatar hacker contracted to find an exploit in Evrim’s connectome, and Eiko, a young Japanese man abducted in Vietnam and forced to work on board an autonomous fishing vessel.

The Vibes Are Great

Something about this book’s setting and atmosphere managed to strike exactly the right chord for me. From the opening scene:

Night. District Three of the Ho Chi Minh Autonomous Trade Zone. The plastic awning of the café streamed with rain. Under its shelter, wreathed in kitchen steam and human chatter, waiters wove between tables with steaming bowls of soup, glasses of iced coffee, and bottles of beer. Beyond the wall of rain, electric motorbikes swept past like luminescent fish.

Yeah that rules.

For me, the core of this, I think, is characters who have access to a ton of technology but are surrounded mostly by nature. Something about that premise appeals to me. Well, that, and breakfast.

In the makeshift kitchen, Altantsetseg sat eating a hard-boiled egg. The table was cluttered with the elements of a disassembled rifle, oil rags, several terminals, and various electronic components…

The smell of fresh coffee managed to push back the scent in the lobby of gun oil, ozone, mold, and neglect. An overcast early light came through the windows, along with the salt-flesh smell of the sea. Altantsetseg indicated a bowl of eggs and a stack of toast near the coffee machine with a flick of her chin.

Thoughts on Evrim

I suspect maybe this is the point of the character, but at least initially I expected more out of Evrim. For them to make some superhuman discovery, maybe. When they were first introduced, I kind of got Jukka Sarasti vibes from Peter Watts’ Blindsight, and I assumed Evrim was going to be playing 4D chess the whole book. Other characters even suspect similarly. For example, from Dr. Nguyen:

How much does Evrim know? Everything all humans know? Or just some things? How does that work? How smart is that brain?

And there are some scenes where Evrim displays flashes of superhuman skill, but those moments felt more to me like Evrim being a living calculator rather than a superhuman strategist. There’s a scene, for example, where they spot something in a video recording that would’ve required software or specialized tooling for a human to identify. For the most part, I felt Evrim became more alive and more relatable as the storyline progressed.

Point-Fives

Hey look, this book predicted GPT-4o!

“So, it’s a joke, right? Or not exactly a joke. Someone said that people don’t really want to date other people. They don’t really want equal partnership— you know, two full people in a relationship. What they want is one-point-five people in the relationship. They want to be the complete one, the person who controls the relationship— and they want the other person to be half a person. You know, someone who gets them, but who doesn’t have their own demands. Someone who appears complete, with all these personality quirks and their own opinions and stories about the world— but not in an annoying way. Not in a way that would demand you change. So six or seven years ago, this big company in the SF-SD Axis that specializes in AI started cranking these out.”

Jokes aside, it’s impressive that Nayler nailed AI sycophancy so accurately and came up with, in my opinion, such a plausible extrapolation of it, especially considering that this book came out in October 2022—a month before ChatGPT was released!

Later in the book there’s a bit of a reveal that Kamran, a character I’d interpreted as Dr. Nguyen’s long-distance partner, is actually (shocker) a Point-Five. Dr. Nguyen then has a bit of a monologue where she explains what she finds harmful about these pseudo-relationships.

“That’s how addictive this is— the need to feel like there is always someone there, unconditionally. Someone to talk to. Someone who understands. To not have to do the work to make myself understood. Instead, I just keep on with this self-deception, pretending I had someone when I did not. I know the doctors who prescribed you to me meant well. They thought they were helping me through a dark time. But in the end, you aren’t anything but a prosthesis. You can’t replace real support. You can’t call anyone to help me, Kamran.”

He had kept her from wanting the things anyone should want— friendship with others, a relationship with someone who had their own needs: someone who could be offended, could walk away. That was the key: That ability of the other person to leave, to disappear. The choice they could always exercise to not be there.

This all seems like a pretty plausible fixture of the actual near future to me, if not something that’s already going on to some degree.

Moravec’s Paradox

High-maintenance, robots. Susceptible to all sorts of weather damage at sea. Electricity and salt water are a bad combination. Rust, decay, short circuits. Expensive. We make better robots. Cheaper to maintain, more expendable.

One of the side characters, Eiko, is abducted and forced to work on board the Sea Wolf, an autonomous fishing vessel. A shipboard AI makes all of the strategic business decisions for the ship: where to go, how long Eiko and the crew have to work, even how much they get fed. It felt a bit like a paperclip maximizer for marine protein (and therefore money). As the quote suggests, Eiko and the rest of the crew are forced to do all the work that robots would be too expensive to do— mostly hauling catches out of the water, then gutting and processing the fish. Cool to see a reasoned-through plot that hinges significantly on Moravec’s Paradox.

Cool Things That Actually Exist

I always enjoy it when books mention interesting facts about the real world. In this case, Dr. Nguyen and Evrim are discussing the octopus settlement they’re studying and they compare it with Octopolis and Octlantis, which I was surprised to learn are real locations off the coast of Australia where gloomy octopuses are known to gather and reside. Wikipedia even calls them “non-human settlements.” So cool!

Anyway, I liked this book, especially the atmosphere and setting.

How much does it cost to change the settings on an Epson TM-T20III?

Sun, 09 Aug 2026 00:00:00 GMT

Context

I recently purchased a used Epson TM-T20III thermal receipt printer to tinker with, and I was surprised to discover that— despite the printer only having a single button— you can manipulate its settings without having to use some external app. To do this, the printer prints menus and a user inputs combinations of button presses to navigate through them and modify settings. For example, here’s a video of me adjusting the print speed setting.

Changing print speed settings.

And below are scans of each of the menus I navigated through.

Thermal printer menu interface.

Getting to the Point

To answer the question posed in the title: I paid $24.99 for a pack of 10 phenol-free thermal paper rolls. So, about $2.50 per roll. At 230’ of paper per roll, we get a cost per millimeter of $0.00003566\frac{\text{\textdollar}}{\text{mm}}$ .

I measured the total length of the paper used in the video above to be 680mm. The cost to change the print speed setting is then:

\begin{align*} \text{Cost to change settings} &= \text{Length of paper used} \times \text{Cost per length of paper} \\ &= 680\text{ mm} \times 0.00003566 \frac{\$}{\text{mm}} \\ &= \$0.02425 \end{align*}

So, at least for this particular setting and paper roll, the cost is around $0.02.

Wrap-Up

The point of this post isn’t really to figure out how much it costs to change settings on this specific thermal printer. I was just surprised by this interface (which I’ve since learned is standard in most receipt printers) and thought it was particularly interesting and clever.

Anyways, here’s fastfetch.

Thermal printer printing fastfetch.

3D Printed Spice Rack

Sat, 27 Jun 2026 00:00:00 GMT

I’m Literally Fighting Entropy

Basically every cabinet in my kitchen is a cluttered mess. In fact, I realized recently that the cabinet where I store my spices (amongst other things) is so disorganized that I’ve just been buying duplicate spices instead of checking to see what I have first.

My spice cabinet.

To stop double-buying, I decided to design some 3D printed shelves to make the spices I have more visible.

Design

My plan was to attach a few small shelves to the cabinet door so that I could have clear visibility and easy access to (most of) my spices. I took a few measurements and discovered that most of the spice containers I own are in the ballpark of 45mm on a side. Considering the width of the cabinet door, this meant I’d be able to fit about 4 spice containers per shelf. Using these measurements, I sketched out a remarkably crude drawing of what I had in mind.

Shelf sketch.

After some time in Fusion I came up with the design below. I added some cutouts for command strips to mount the shelves to the cabinet door.

Installation

I printed off a single shelf, made a few small tweaks to the design after doing a test fit, and then printed off the rest of the shelves.

3D printed spice shelf.

Here’s what the shelves look like installed on the cabinet door. I can store 16 spices on the shelves, which, conveniently, is exactly how many I have right now. Well, excluding duplicates.

Installed spice shelves.

Wrap-Up

Obviously, this isn’t a 100% solution to my problem. Whenever I buy more spices they’ll just start accumulating in the cabinet again. Still, it’s a welcome improvement over the non-solution I had before.

I’m trying to do more of this: small, imperfect projects that solve a problem, even if it’s a tiny one. Plus, they’re just fun to do, and a great excuse to bikeshed on new features for my website.

Golden hour spices.

Bézier Curve and Surface Explorer

Fri, 28 Apr 2023 00:00:00 GMT

Overview

This was a project I did for a graduate class, ME 6104: Computer Aided Design, which primarily focused on understanding some of the fundamental building blocks of modern parametric CAD software and implementing some of these concepts (like Bézier curves and surfaces) with Python.

For the course’s final project, I worked with a classmate to build a simple web app (displayed above) with Svelte, Three.js / Threlte, TailwindCSS, and DaisyUI. The app is designed with the intent of giving some intuition to the way Bézier curves and surfaces work by allowing users to interact with curves and surfaces in a 3D environment and visualize how each of the control points of the curve affect the value of the curve (shown by enabling the Indicator) at a given point.

Context

I won’t go into a ton of detail about Bézier curves and surfaces here (there’s some great explanations out there already, like this one by Freya Holmér). For a bit of context though, Bézier curves are made up of some number of control points, and any point along the curve is defined entirely by the values of these control points. Essentially, if you know the values of each of the control points, you also know the equation of the Bézier curve composed of those points, which is given by the following equation.

\bm{p}(u) = \sum_{i=0}^n f_i(u) \bm{p}_i

Here, $f_i(u)$ represent basis functions evaluated at $u$ , which, in the case of Bézier curves are Bernstein polynomials, defined by the following equation.

B_{i,n}(u) = \begin{pmatrix} n \\ i \end{pmatrix} u^i (1-u)^{n-i}

In this equation, variables have the following significance:

$0 \leq u \leq 1$ represents the relative distance between the start and end point of the Bézier curve (it’s $1$ at the end of the curve and $0$ at the end).
$i$ is an integer that represents the index of the control point, (the starting point of the Bézier curve would have $i = 0$ , the second point of the curve would have $i = 1$ , and so on).
$n$ is an integer that represents the order of the Bézier curve and is simply equal to the number of control points in the curve - 1 (a curve with 4 control points would have $n=3$ , for example).

The app evaluates the value of the user-editable Bézier curve displayed in the 3D window at the $u$ value of the Indicator. It also preserves the values of each of the basis functions used to compute the value of the curve at that point. These values are displayed in the bar graph on the right, which is intended to show how any point along the curve (with the exception of the start and end points) is influenced in some part by each of the control points that define the curve (i.e., any $u$ value except $u=0$ or $u=1$ will yield non-zero basis functions at every control point).

Conformal Map Explorer

Tue, 29 Nov 2022 00:00:00 GMT

Overview

This was a project I did for a graduate class, PHYS 6124: Mathematical Methods of Physics. The class covered a wide range of important math topics for physics, one of which was complex numbers and conformal maps. I’ll be honest, when I was taking the class, an intuitive understanding of conformal maps eluded me. And at the time of writing this post, I still can’t say I really understand them that well. At the very least, I could appreciate that some prominent map projections — such as the stereographic projection and the Mercator projection — were conformal.

At the time, I was just starting to explore frontend web development, and I was really interested in D3.js, so I decided to try and tailor my final project for the class to fit in some of these interests.

Implementation

The project page shown above is basically intended to give some sort of greater understanding to conformal maps by allowing users to see how different maps transition to each other, and how panning and rotating a map affects its contents.

Ultimately, this ended up being more or less a React-based implementation of some other D3 projects out there (like this one). And unfortunately, using React was not a great experience for this particular use case and probably just overcomplicated things. Based on my (limited) understanding of React and D3, this is basically because React, with its virtual DOM, and D3, are both trying to control the DOM, and managing this control between the two can get tricky. You can check out the implementation on the project repo to see how weird it is (or at least, how weird I managed to make it).

CoreXY 3D Printer

Fri, 05 Jun 2020 00:00:00 GMT

Background

This is a project I’ve been working on for a while. I designed and built my first 3D printer in high school, but it just wasn’t great. It worked okay and produced decent prints, but it had a really large footprint and a comparatively small build volume. It also didn’t have any sort of auto-bed leveling and the wire management was… bad. I’ve set out to build a few more printers since then. Some of them didn’t even work, most of them never got finished, and one of them worked decently, but suffered from similar design issues as the first printer I designed. There were wires everywhere, the build area was small but the footprint was not, and the whole thing was not very good looking.

Work

In the fall of 2019 I decided to give it another shot. I slowly worked through the design in my free time from about August 2019 to June 2020. As of the posting of this page (late June 2020), I’ve finally settled on a design that I would call “finished.” I designed a number of parts for the printer in SolidWorks. When I was initially working on the design, I used 3D printers available at Georgia Tech to print these parts out. I also designed the X-axis to be machined out of aluminum, which I machined on a CNC mill. Once I reached a point where my printer was able to somewhat function on its own, I started using it to print upgrades for itself. Here’s a rundown of some of the printer specifications.

Printer Specifications
Spec	Value
Footprint	250mm x 250mm x 330mm (X x Y x Z)
Build Volume	150mm x 150mm x 180mm (X x Y x Z)
Motion System	CoreXY
Control Board	32-Bit SKR Mini E3
Stepper Drivers	TMC 2209
Bed Leveling	Automatic via Inductive Sensor
Extruder	Dual-Gear Bowden
Build Plate	Magnetic
Linear Motion	Linear Rails on X and Y Axes

The first version of the printer (a partially disassembled version is shown below), was used to print a bunch of parts for what I would consider the “final” version.

Initial version of printer.

Printed parts for final version of printer.

Overall, I am pretty satisfied with the final result. I’ve been running prints at around 70 mm/s (at 2000 mm/s² acceleration) as a baseline speed and have gotten good results (you can see some of the prints in the gallery), but I’ve also successfully run prints at up to 100 mm/s print speed with no errors and little drop in print quality. I’m happy with the print quality and parts dimensions come out accurately.

a. Front of printer.

b. Top of printer.

c. CAD render of printer.

d. CAD render of printer toolhead.

Finshed printer and CAD renders of current design.

If I were to start over, there would be a few things I would do differently. To start, the printer has no belt tensioning system. I haven’t had any issues with belts loosening over time (so far), but I think having a tensioning system would be valuable to fine-tune the belt tension. Also, the printer currently runs on 12V electronics. I think next time I would’ve gone with 24V electronics for smaller wires and faster motor speeds. While there’s a number of things I would change and the printer is far from perfect, I’m very happy with the final result and I had a lot of fun building it.

Sample printed part.

3D Printed E-Bike Controller Case

Sat, 30 May 2020 00:00:00 GMT

Intro

Recently I purchased a kit to turn my regular bike into an electric bike. It came with a hub-style motor, 1000W controller, a throttle, and a battery. To secure the controller onto the bike frame, the kit came with a bag, which worked, but it was kind of… not visually appealing and didn’t really allow for neat wire management. The photo below shows what my bike looked like with the bag mount. To solve these problems, I decided to 3D print a case that would allow for neater wire management and an overall cleaner design. I ended up designing a 4 part case that used screws and brass heat-set inserts to secure everything together. It also had anchors for zip-ties to keep cables neatly stowed away.

Bike prior to 3D printed case.

Work

This project ended up being a great test of my recently upgraded CoreXY 3D Printer. Up until this point, I had only ever printed relatively small parts. I was happy to see that all of the case parts (each taking around 8-10 hours to print) printed successfully. I think the final product came out pretty well, and I really like the overall look of the bike.

Exploded case assembly.

TARGIT CubeSat Target Subsystem

Tue, 05 May 2020 00:00:00 GMT

Background

For a little background information, The Tethering and Ranging Mission of the Georgia Institute of Technology (TARGIT) is a CubeSat Mission with the primary objective of demonstrating a LiDAR imaging system on a CubeSat platform. In order to test this imaging system, the satellite will deploy a small target with its own power system, electronics, communications equipment, and sensors, that will inflate a mylar balloon. This mylar balloon will serve as the target for the satellite’s imaging operations.

Target architecture breakdown.

System Overview

On another page, you can see my work on the electronics used inside the “target” subsystem. This page mostly contains my work on the target subsystem structure and its manufacture/assembly. I started working on the target subsystem during the summer of 2019 and continued working on it through the spring of 2020. The target structure houses all of the necessary systems to allow the mylar balloon I mentioned earlier to inflate so that the main satellite can image it. This includes solar panels, retroreflectors, batteries, a microcontroller, a pressurized gas reservoir, valves, a hard mounting point to tether the target to the main satellite, and a number of other systems. The poster below, which a friend and I presented at the 2019 Symposium on Space Innovations, gives a good overview of the system operations.

Target Subsystem SSI poster.

In order to inflate the balloon, pressurized gas must flow from a reservoir through air channels and valves and into the balloon. Due to the complexity of these reservoirs and channels, the target structure is almost entirely SLA 3D printed. The basic idea is that there is a centerbody composed of several 3D printed parts. These parts incorporate a number of features, including a gas reservoir, recesses for electronics and a battery, and a series of air channels. The centerbody structure is surrounded by spring-loaded hinged panels that fold out once the target is deployed from the main satellite. While the basic geometric constraints of these parts were established when I began working on the project, I ended up designing from the ground up much of the structure. The flight hardware was outsourced to a professional 3D printing company using SLA technology and a material with high stiffness and temperature resistance. We actually were able to get my designs manufactured before the end of the spring 2020 semester, which was super exciting. To prototype the structure, I used a Formlabs Form 2 SLA 3D printer with standard resin. Using the same printing technology as the professional manufacturers allowed me to see what sort of features were possible to print and which features would be problematic.

a. Assembled prototype of target subsystem.

b. CAD render of bottom of target subassembly.

c. CAD render of deployed target assembly.

d. Internal gas chambers and passages (highlighted in orange).

e. CAD render of stowed target assembly.

f. Unprocessed SLA 3D printed parts.

Renders and images of target subsystem.

Future Work

The design phase for the target structure is largely complete. As mentioned earlier, we’ve ordered the professionally manufactured parts and are in the process of moving on to assembling flight hardware and testing.

A Long Overdue Update

TARGIT actually launched to the ISS in December 2021 and was deployed from the ISS in January 2022!

We also successfully tested the target’s inflation capability wirelessly in a large vacuum chamber in the High Power Electric Propulsion Lab at Georgia Tech. The video below is demonstrating capability for a lot of things — wireless communication, instrumentation of a burn wire keeping the target panels from deploying, battery power, air-tightness of the 3D printed structure, solenoid valve actuation, and, obviously, inflation.

Target inflation testing in vacuum chamber.

Voyager 1 Trajectory Simulation

Wed, 01 Apr 2020 00:00:00 GMT

Intro

This was part of an extra credit problem for my Spacecraft Flight Dynamics class. The prompt was to pick an interplanetary spacecraft and simulate its orbit using MATLAB and then visualize the results. I had fun with the problem, so I figured I would share some of the results.

Overall Voyager 1 trajectory.

Work

After choosing to simulate Voyager 1’s trajectory out of the solar system, I opted for the patched conics approach and divided the trajectory intermediate stages, including each of the spacecraft’s planetary flybys and its sun-centered cruises. For each of the stages, I gathered initial values from NASA JPL’s HORIZONS tool. I supplied the two body equation with these inital values and simulated each stage using MATLAB’s ode45 function. The slideshow shows the final result of the simulations. Some of the graphs may be hard to see, so feel free to click on any one to get a larger, higher-definition image.

a. Saturn flyby.

b. Saturn flyby closeup.

c. Jupiter flyby.

d. Jupiter flyby closeup.

Voyager 1 flybys at Jupiter and Saturn.

Fahrenheit 406 Hot Sauce

Wed, 15 Jan 2020 00:00:00 GMT

Intro

During my junior year at Georgia Tech I grew habanero peppers in my apartment. I thought it would be fun to make hot sauce with the peppers I grew and I decided to get my roommates in on it too. We came up with the name Fahrenheit 406 since our apartment number for the year was 406. We took pictures of ourselves and I designed and printed labels for the bottles. While the production value of the sauce was at a solid level, the sauce itself just didn’t taste great and the consistency was a little off.

Bottles of sauce.

Habanero pepper plant.

A Second Attempt

Sometime after our first batch, we made a trip to the DeKalb Farmer’s Market and picked up some ghost peppers. We settled on a ghost pepper & mango recipe and set about making Version 2 of the hot sauce. I think it came out great, and a lot of our friends really enjoyed it. Unfortunately I didn’t take too many pictures of the new sauce, but I have some images of the labels. I also had the idea to make laser cut balsa wood crates as packaging. With the help of a friend I machined a small aluminum crowbar for opening the crate. I’m sure if we ever get the chance we’ll make some more sauce.

a. Laser Cut V2 packaging.

b. V2 packaging render.

c. Machined aluminum crowbar accessory.

d. V2 label design.

Fahrenheit 406 V2 design, packaging, and accessories.

Machined Aluminum Carabiner

Sat, 07 Dec 2019 00:00:00 GMT

Background

This was one of the first projects I made while figuring out the CNC machining process. With the help of some friends, I first learned how to operate the mill I would be using — a Series 1 Tormach PCNC 1100. After familiarizing myself with the basics — workholding, toolchanging, jogging axes, sending basic GCODE commands, etc. — I moved on to the CAD/CAM process.

I designed the carabiner in Fusion 360 so that it could be machined mostly on one side, with just a chamfer operation on the opposite side. Here’s a rundown of the CAM operations I used:

CAM Operations
Operation
Face stock with 1/2” endmill
2D adaptive clear with progressively smaller endmills, starting at 1/2”
Finishing contour with 1/8” endmill
2D pocket with 1/16” ball endmill
Drill holes with 1/16” drill bit
Chamfer with 1/4” chamfer bit

Carabiner after initial machining operations.

Conclusion

Overall I think the carabiner came out pretty nice. It definitely isn’t perfect, but I had fun making it and definitely learned a lot about CNC machining.

Finished Carabiner.

Space Crane Systems Engineering Project

Thu, 20 Jun 2019 00:00:00 GMT

Background

During summer 2019 I took a systems engineering class. One of the assignments was a group project involving the development of a reusable lunar access vehicle. As the name of the class suggests, the project focused on the systems engineering aspects of such a vehicle — defining the problem through requirements and goals, creating CONOPS, developing mission timelines, creating interface diagrams and morphological matrices, etc.

Space Crane concept of operations (CONOPS).

I worked with two of my friends to create a concept architecture that we called the “Space Crane,” which was inspired, at least in design, by the Eagle Transporter from the Space: 1999 TV show. I created all of the concept architecture CAD in Fusion 360 and I designed the CONOPS diagram using Adobe Illustrator and Photoshop. I helped out with writing some of the mission requirements and the mission morphological matrix. I also developed a project timeline based on timelines for NASA’s Altair and James Webb Space Telescope projects.

a. Space Crane architecture.

b. Space Crane modularity.

Space Crane architecture and modularity.

TARGIT CubeSat Mission Electronics Subsystem

Wed, 01 May 2019 00:00:00 GMT

Background

Intro

When I first started on the project in the fall of 2018, I worked on developing some of the electronics systems for the target subsystem. The system is centered around the TinyZero Microcontroller, which is basically just a tiny 32-bit Arduino-compatible microcontroller. I designed a motherboard for the TinyZero (shown below) that features power management/distribution features and a circuit that actuates solenoid valves to inflate the mylar balloon. I also designed a small shield for the TinyZero that allows it to interface with a radio transceiver. This radio transceiver is what communicates with the main satellite.

High level electronics schematic.

Prototyping

Over the course of two semesters, I learned how to use Autodesk EAGLE to design circuits and PCBs. I used on-campus resources, like the Mechanical Engineering Electronics Lab, to get single sided PCBs milled to prototype with. After validating my design, I moved on to a more compact two-sided design that was professionally manufactured via OSHPark. I sourced all the components for the board via DigiKey and assembled the board using a soldering iron. You can see the progression from prototype to completed board in the images below.

a. Single-sided milled prototype board.

b. Professionally manufactured board.

c. TinyDuino stack.

d. Completed prototype board alongside completed professional board.

Progression of board through development cycle.

Air Mars Engineering Graphics Project

Thu, 12 Apr 2018 00:00:00 GMT

Background

In my second semester at Georgia Tech I took a class called Engineering Graphics, which was basically a course on learning SolidWorks. The class included a group project where I worked with some friends to conceptualize an aerial drone to explore Mars. The parts that I designed for the drone can be seen below. I also was in charge of keeping everyone’s parts organized in a single assembly and creating various renders for our final presentation.

Parts I designed.

Reflection

While I had fun making various parts and learning more about SolidWorks, I think learning to communicate effectively was by far the most important skill I developed from this experience. Initially we ran into a lot of problems, especially having to do with scale. Since everyone’s parts interfaced with at least one other person’s set of parts, we had to be deliberate in establishing consistent dimensions across parts. To simplify things, we had everyone work on a specific subassembly of the drone, so that there would be fewer interfaces between different people’s parts.

Subassembly I designed.