Electronics on Arturo Negrette

Chronicles from a Film Lab Tech

Thu, 11 Dec 2025 14:50:00 +0000

Background

I’ve been working at a film lab for the last couple of months, and the engineering degree immediately made itself useful by giving me the tools to keep some of these machines alive. Problems ranging from just corrosion within a sensor’s connector to open heart surgery on a board no longer made. For a glorified sous-vide tank, the Noritsu engineers put a lot of safeguards and smart engineering in place to keep these working as well as they have for as long as they have. I’m writing this article with the intention of logging whatever I find for whoever takes over for me, or whoever finds this website in similar despair as I have googled whenever something broke.

This is the V50, relatively sophisticated MiniLab unit for C-41 Color Development. Short Leader style, which means you tape the film to a sheet of plastic with sprocket holes on it, and that’s what drags the film itself out of the can and into each step of the development process. The 50 in V50 stands for 50 rolls an hour, so not too shabby accounting for the fact a 36 exposure roll of film tends to be somewhere around 5 feet.

In order to put the rest of the machine in context, let’s cover the development steps of Black and White and C-41 film:

For both, the film is coated with silver halide crystals. Exactly how they are spread on the film and what they do during and after the development process varies between B/W, C-41, and every other development process available, but the working principle remains the same. After being exposed to light, these crystals are raised to a higher energy level, leaving a latent image stored chemically on the film emulsion.

B/W:

Developer: This step converts the latent image contained in the activated silver halide into metallic silver. This metallic silver only appears in the locations where light struck the emulsion. The stronger the light, the more concentrated the silver, the darker or "denser" the image is at this location, and the reverse applies as well. All of this to say that we now have a negative image of what was captured in the camera at that moment.
Fixer: After the developer, the image that matters to us is created. However, it's still surrounded by the inactive silver halides still in the emulsion. No light, no developer action, no metallic silver. The fixer can dissolve the halides but not metallic silver, which allows us to selectively remove the unused stuff and keep our image negative.

Easy stuff, we’ve been doing monochrome stuff essentially as long as we have been able to understand organic chemistry. B/W film made a year ago and 80 years ago work the exact same, and excluding the effects of age, can be developed the exact same way. Color, however, is where modernity truly shines through. Kudos to all the chemists and engineers that allowed us to record the world around us in more than shadows and highlights.

C-41:

Developer: Still does the same as B/W, with the caveat that now the dye couplers (part that creates the colors) are activated and create the dyes of each layer. Fun fact, you can cross-process any C-41 roll as B/W, you just ignore the color couplers for a monochrome image. I've managed to salvage pictures from obsolete C-22 rolls from 60+ years ago. Hard drives fail, files go corrupt, but in a decently wide range of conditions, chemistry prevails. Cool stuff.
Bleach: Due to the layout of a typical C-41 film emulsion, after development, you now have both a dye-based image, and the metallic silver image. Keeping both of these leads to super dense images with low saturation, and often it was used intentionally as a color grading effect in movies and TV in a method called "bleach bypass". The bleach attacks this metallic image and converts it back to silver halides, which can now be removed by the fixer. It also helps reducing the density of the dyes in the film base.
Fixer: Same as B/W. Fixer actually holds quite a decent amount of silver, so it's very important to dispose of properly and feed it to silver reclaiming facilities. Not that any other chemical is great for carelessly dumping into the environment, but the fixer carries metals with it that are specially bad to introduce into an ecosystem.
Stabilizer: The dyes contained on each layer are technically not fixed in space, and age, environmental factors, and improper handling of the film can lead to them migrating between layers. The end effect is a reduction in the color reproduction quality, so stabilizers are used to "lock" the dyes out of escaping their layer. You can very well skip this step if you are just looking for a glorified Instagram filter, but in order to produce negatives that can withstand the test of time, stabilization is crucial.

I’d love to get into detail about the careful design of color emulsions, but so have another 500 trillion people on the internet, so I will unfortunately not elaborate here. Google it, watch Tech Connections about it, don’t use this case of brevity as an excuse to avoid learning about film.

Our machine feeds the leader card and the film it’s attached to into a very deep 18 Liter Color Developer (CD) tank at a speed that allows for each length of film to be in contact with the CD solution, while automatically adding a replenishment amount of fresh chemical to account for exhaustion per roll. All the chemicals involved in these reactions are disposable and need to be replenished accounting for the surface area that passes through. The film is continued to be carried in and out of the solutions automagically by careful mechanical design of roller transports like conveyors through a factory. The machine doesn’t actually know where this film is, but it vaguely knows when it started being fed and when the canister was cut off at the end, and it knows how fast everything is spinning and moving the film along. So it starts counting and can make an educated guess as to where it’d be and what chemicals would be exhausted at which times.

At the end, it passes through essentially a gentler air fryer before being released fully dry for it to be scanned and/or sleeved for preservation. Almost magical coming from my background of hanging up film from my shower curtain loops.

I’ve neglected to mention much about the modified V30 that develops B/W, but it’s the same principles and same tech, just behaving differently to accomodate for the different chemistry. Both operate the same. Now, let’s get into it.

Repair Log

Most Recent: B/W Intermittent “Bootlooping”

Symptoms

Machine goes into a cycle of booting and self-testing, encountering some error and powering down, then repeating the cycle.
Initially only happened at the end of a roll, right after cutting the canister off. One reboot cycle and it would continue as normal.
Progressed to continously turning off and on again.

Troubleshooting + False Solution

My initial assumption was an intermittent short in a system related to the springloaded blades that cut the end of the film from the canister. It’s a guillotine that’s tensioned until it hits a detente to hold the spring securely. When the can is pulled closer as the roll runs out, it hits a paddle that trips the spring catch and the blade flies into the film, cutting it off. The spring is loaded via a motor on a cam, so I started checking that the motor actuated freely, which it did.

At the beginning of each day, a start check is completed where a leader card is fed through the machine with no film attached to it in order to confirm that there’s nothing wrong with the extensive tree of gears and rollers that transports everything from beginning to end. I noticed that during this test the problem still occured, even though the motor was not involved and the cutter was not triggered. It ruled out anything wrong with that subsystem, giving me a dead end.

Visiting the service manual pointed me to a logic issue with the front panel control board, where I started poking around until I found out through a tech at one of the other labs that something similar happened when the front panel button connection was corroded. Upon cleaning and reseating the cable the problem was fixed. Straight forward enough.

A couple of months passed and the problem returned, where we’d clean the cable and reseat it, helping for a couple weeks until the issue appeared again. Every cleaning would return everything to normal but the time until the problem reappeared was slowly getting shorter until now.

Solution

At this point, it had been several months of increasingly more regular cycling on the ribbon cable. The usual cleaning and reseating got to a point where it did not solve the issue at all, the machine could not be used at all. When I took it apart and inspected the ribbon cable under a loupe I found that the contact material had fully worn through, and I presume it wasn’t making contact at all. I unfortunately did not take a picture of the damage before the first fix, but looking at the picture below, the untouched side shows the gist of it. The contacts on the board wore a hole through the pads.

Version 1

We were pretty backed up in terms of black and white orders, so I needed to get the machine to stop cycling and begin processing film. I had the idea of using aluminum tape from one of the film stocks we sell. They make their DX codes by sticking some aluminum tape and putting a label over it in order to make the pattern of conductive/non-conductive squares. Pretty neat. Turned out pretty sloppy but it got the job done to catch up with orders. Let’s call this one Version 1.0

Unfortunately, some of the buttons were intermitently working, so I knew we couldn’t keep this for the long term. We had ordered some X-Acto knives so I figured I’d redo the patchwork with sharper blades whenever those came in. Version 1.1 turned out more aesthetically pleasing but actually gave its own set of issues. For example, the set of buttons that were acting up with 1.0 returned to life, but a different set did not work at all and in fact ended up sticking and led the machine to behave erratically. This told me that I was now shorting adjacent pads to one another or to ground around the connector body. At this point I realized the tape we use for taping the film to the leader cards was pretty perfect as a low-temp Kapton tape alternative and the idea for V2 was born.

Version 2

For V2 I had two major conclusions driving how I went about this next repair. First, I needed to shield everything besides the actual contact point in order to prevent any more shorting. I was very lucky I didn’t fry anything that last time. I also needed to have better contact between the original flex PCB traces and the aluminum tape, so I needed to flip the aluminum tape in order to keep the glue from acting as an insulator.

By stenciling the actual location of the worn contacts onto some tape, I was able to shield everything else and it made the size, shape and location of my aluminum tape squares less critical. I was also able to stick them onto the bottom of the tape and rinse off the adhesive from it with some alcohol. It took me the longest out of all the solutions I came up with but at this point it’s the closest I’ve gotten to a permanent fix short of replacing the board altogether.

Color Developer Working Tank Overheating

Symptoms

Intermitent S-Sol Thermo Activated Error
Solution reaches 43C before other safeguards kick in
Initially only happened during overnight sleep cycle

Initial Troubleshooting

Compared thermocouple readings to thermometer in working tank and in side tank (will elaborate on tank layout in a second) – Temp Matches
Drained tank, cleaned agitation and recirculation pumps – Minor to Negligible Improvement

For context, we had just recovered from the below problem of the cutter jamming, so at this point we had been chasing problems around for around a month.

Preventative maintenance is crucial for keeping old hardware like this up and running. Much like one of my old bosses used to say: “Schedule the maintenance or the machine will schedule it for you”

Going back to the problem, at first the heater would behave but over time it started getting stuck powered on during normal operation and not just during the overnight sleep mode. When we confirmed that the machine was reading the correct temperature for the tank, it indicated to me that a relay was getting stuck closed intermittently. Still, it’s a team project and the rest of the team wanted to confirm nothing else was out of the ordinary before ruling in favor of swapping the relay. I didn’t work for a couple days so the rest of the team limped the machine through and only took 3 day turnaround orders.

I took a quick brush through the service manual and found that the stabilizer control system and the one for the tank giving us problems was the exact same, so I swapped its thermocouple and heater, making the machine believe that the stabilizer was the color dev tank and viceversa. The heater loops all were controlled via components labeled SSR1 through 7, all on the power board at the bottom of the machine. Unsurprisingly, the stabilizer was now overheating instead. Luckily, we could run the stabilizer at room temperature so disabling the color developer heater could let us run the machine and catch up with orders before a repair needs to be done. I reset the thermostat for the stabilizer to what it needed to be for color developer (remember that color dev is now stabilizer from the point of view of the machine and viceversa), and then set the color developer temperature to zero so it would not attempt to power anything on that circuit. The machine slowly got up to temperature and remained there as it was now able to turn off the color developer heater.

At this point we caught up with orders for the most part, so I carried on with trying to fix the root cause. When I came back to the machine, I noticed that although the troubleshooting LEDs on the power board showed the control unit was not powering the heater relay, temperature continued to climb. From my car background I had the intuition to begin tapping around every black relay-looking box with the rubber back of a screwdriver. It didn’t help.

Turns out SSR stands for Solid State Relay, something I realized pretty late, embarassingly late. SSRs tend to fail closed as they’re semiconductor devices, and it confirmed and explained my initial assumption of a stuck relay. The awesome engineers at Noritsu had the foresight of including redundant relays on the board, or rather, populating relay footprints for heater circuits not used in this machine model. I’m sure the V100 or the V5Billion use these relays, but I’m lucky mine doesn’t because these exact P/Ns are looooooong discontinued. In my panic before realizing there were spares I did find a suitable drop-in replacement. I crossreferenced the info on the part itself, the pin pitch from the board and the specs from the service manual. Part number is D2W203F-11 and below are some links to find a replacement in case you’re not so lucky to have extras:

Digikey Mouser

Closely inspecting the board would’ve saved a bit of time, but it wasn’t easy to get out of the machine and I didn’t want to take it out unless absolutely necessary. These things are a couple grand each.

A candidate for the world’s oldest solder was holding the relay in place, so a ton of flux, time and patience were needed to get it out and replaced with the extra. The pad was fried to begin with, and I presume that solder started cracking and creating a high resistance connection. The thermal cycling probably also weakened the joint and created a feedback loop culminating where we are now.

After it was all said and done, I gave the board a good dusting and a rinse with isopropyl alcohol to wash of the flux all over the board and give it some TLC. Once installed, the heater problem was solved and it’s still powering through orders to this day.

Cutter Jams

Symptoms

Right Lane Struggled to Cut Canister at the End of the roll
Canisters stuck after cutting
Cutter Return Error

Initial Troubleshooting

Initially,

Film Chemistry Experiments

C-41 Reversal

Black and White Reversal

Project SABR

Sat, 03 May 2025 14:50:00 +0000

Major Achievements:

Developed a fully-fledged testing platform for air-breathing RDEs at UCF
Laid the foundation for a future senior design teams
Produced data for several AIAA SciTech 2025 papers

Background

Rotating Detonation Engines (RDEs) are a relatively novel form of rocket propulsion that takes advantage of pressure-gain combustion. Conventional rocket engines use constant-pressure combustion cycles. RDEs theoretically produce better performance by taking better advantage of the energy stored chemically in the propellants. There is a fine line between detonating in the way that yields the performance advantage, and detonating in the way that destroys the engine. Due to the speeds involved in the process, these engines have only recently began gaining traction now that computers and sensors allow for the extremely fast sampling rates necessary for reliable data output. Their compact size and efficiency gains allow for storable propellants to be used more effectively in missile and defense applications.

The goals of this project at the beginning were beyond creating a working air breathing lab scale engine, to lay the foundation for further research to be done at UCF. This involved creating a test stand equipped with a capable and robust data acquisition system and control system. Considering the tight budget for a project of this ambition, it took a lot of creativity and clever thinking to get it off the ground (metaphorically).

Design

Ignition

Lighting an RDE involves precise timing and a strong enough relese of energy to induce a detonation mode within the engine. A fast burst (in tens of micro seconds) with not a lot of energy is more likely to induce a detonation compared to a longer burst with a ton more energy. Typically predets (short for predetonator tube) are used where an explosive mixture of gases are lit on one end of a long tube. Over this length of tube, the energy release is structured in such a way a shock is developed by the time the combusting gases exit. Due to the mechanism of operation, these have careful geometry that is hard to package into a system with compact and flight in mind. This is where my approach comes in, conventional torch. If the annulus can act as the constriction that focuses the energy towards a shock, the igniter can be more compact and allow for smaller designs for flight vehicles. After all, the annulus needs to already be able to contain and direct the energy to keep the detonation wave circling around the cavity.

I was in charge of both mechanical and electrical aspects of the ignition system. In terms of the mechanical side of things, I prioritized cost and ease of manufacturing over anything else, arriving at what is essentially a modified stainless steel 1/4 NPT cross fitting, as pictured below. The torch body is then connected to the rest of the RDE via a thick-walled stainless steel pipe nipple. I truly didn’t want the torch to be plagued by scope creep, so it incorporates as much stuff off the shelf as I could find.

With this in mind, the electrical system employs a standard car ignition coil, the cheapest I could find on Amazon. The circuit puts a 12V SPDT relay in purgatory by switching it rapidly between powering its own coil and pulsing the primary of the ignition coil. The best analog I could come up with was those “useless machine” desk toys where turning on the switch powers on the arm to turn itself off. The relay coil is wired to 12V via its Normally Closed contact in series with a momentary switch. When the switch is closed, the contact opens, powers it down, which allows it to return to rest and keep cycling.

Additionally, I added a couple capacitors and resistors finetune exactly how long the charging and discharging takes in order to ensure the primary winding of the ignition coil has time to fully saturate before the power is switched off and the fields collapse to generate the high voltage out of the secondary. I could’ve gone with something based on a 555 timer or even more fancy, but circling back to that thought about scope creep, I truly just wanted it to work and work without needing to think about it. The only drawback is that mechanical relays really don’t like being switched that fast, and because the coil within the relay is also an inductor, there’s plenty of arcing when it’s disconnected during the cycling. Adding a flyback diode across reduced the switching frequency so I arrived at the conclusion that relays are cheap enough to be consumable. During testing, the poor little relay did not fail or weld itself shut once, but if I were to be pulled back to the drawing board for a proper vehicle-born ignition system, I’d probably choose to go with something solid state or capacitor-based much like modern cars use for their sparking.

Your browser doesn't support HTML5 video. Here is a link to the video instead.

A cheeky extra feature I added to the torch is the presence of a pressure transducer line. During startup, the transducer would report the torch internal pressure, but once it’s powered down, it’d act as a CTAP (Capillary Tube Attenuated Pressure) readout for the annulus conditions. The transducer is coupled to the chamber via a 1/4" stainless steel tube to thermally isolate it from the extreme conditions during both startup and steady state operation. The length of tubing necessary was determined via the following function:

\[ \begin{aligned} \frac{T-T_{\inf}}{T_{0}-T_{\inf}}=\frac{1}{\cosh(\sqrt{\frac{4hD_{1}}{12k(D_{1}^{2}-D_{2}^{2})L}})} \end{aligned} \]

where:

$T$ = Maximum tolerable temperature at point of interest
$T_{\inf}$ = Temperature of environment, taken as 100F for this analysis
$T_{0}$ = Temperature of pressure source of interest, in this case, the estimated combustion temperature
$L$ = Length of tubing, in inches
$h$ = Heat transfer rate from tube to still air by natural convection, in BTU/hr/ft^2/ºF
$D_{1}$ = outside diameter, in inches
$D_{2}$= inside diameter, in inches

This model assumes the thin walled tube to be a fin following conduction and convection, much like a heatsink fin. The fin is expected to lose heat by convection at its end, and the pressure source is assumed to only transfer heat via the tube wall and through the pressure media. Another major assumption is that due to the thermal conductivity coefficient difference between the pressurized gas and the metallic walls of the tube, that the heat transfer via the stagnant gas is negligible. This leaves us with a tube length of around 8 inches. I ended up rounding up to 12 just in case we found anything extreme during testing and because the transducer was borrowed and VEEERY expensive. Please bear with me and the demonic US customary heat transfer calculation. I just like inches :^)

In the end, and because our entire operation relied on this very experimental approach working first try, I designed around being able to swap the torch out for a predet and still be able to use all propellant and control connections as normal. By keeping it as a drop-in replacement, we could quickly fall back on something that’s known to work properly and carry on with testing the combustor. There are so many unknowns in RDEs due to the technology’s low TRL that being able to rule out variables saves plenty of long nights of testing at the lab.

As expected, the concept of lighting the RDE was a lot easier said than done. I go more into detail about the outcome under Testing, but the combination of a tight testing timeline, flawed assumptions in the annulus design, and the very experimental torch approach led to us switching it out to try and diagnose the combustor directly.

Structure

Before pivoting towards my work in ignition, I helped with the first handful of iterations of the combustor. I focused on packaging and routing alongside my buddy Hunter, essentially figuring out how to efficiently route propellants from their respective ports coming from the fluids system and into the injectors.

Initially, we weighed the our options for manufacturing methods. We could push for an additively manufactured design, which would allow for complex geometry and reduce the number of critical seals to worry about, but had the drawback of very money, timeline, and risk intensive. On the other hand, we could go for something doable with the resources available at the UCF Machine Shop: reliable and free access to 3 and 5 axis machining, turning, EDM hole drilling and wire cutting. All we’d have to source is our required material. Regardless of whether we printed it or machined it conventionally, the RDE would be exposed to extreme conditions surrounding the detonation wave as it propagates, which means high pressures and temperatures sweeping around in a circle. Compared to a conventional engine, the cyclic failure modes began rearing their ugly head in all of our analysis.

The only saving grace was that the timescales for all this energy release were minuscule, although they compounded over time. Based on the fact we were targetting a very short runtime of only a couple seconds (and much more engineering analysis I will not go into here), we decided to go for 304 Stainless Steel. The reasoning came down to the fact we wanted the ability to fail and iterate without being dead in the water trying to source another brick of material to machine or a whole new printed part. Much of the research behind RDEs is uncharted territory, so even with all of our calculations and analysis double, triple, and quadruple checked, we wanted a large margin of safety on top of the ability to return to the drawing board and modify our design without crippling the project. So with all that being said, we settled on a conventionally manufactured combustor. From all the discussion, preliminary analysis, and talking to researchers both at UCF labs and across the US in places like Purdue and Johns Hopkins, we went for a modular design based on a stack of plates. Below are some of the first iterations of our combustor, initially by me and later modified by Hunter.

A great start, but with any engineering project, we quickly started spotting flaws in terms of component integration, manufacturing, and especially in terms of final assembly in the stand. Looking back while writing this knowing how the project turned out as a whole makes the original design look extra alien. We moved away from fastening the inner body via a huge nut in the back, moved towards longitudinal fluid connections for both of the propellants, and most importantly, designed towards being able to fully assemble the combustor without needing the thrust plate. This last detail made it infinitely easier to mount and remove the combustor onto the test stand during testing. I truly don’t know what horrors we would’ve gone through if we needed the thrust plate involved while putting together everything else.

In the end, the combustor consisted of an inner and outer body, nozzle cowl and spike, and the injector plate and backing plate. The inner and outer body constrain the annulus dimensions and their negative space form the cavity where the combustion occurs. The inner body, however, also acts as part of the internal plenum for one of the propellants. The selected injector schema is “Jet-in-Crossflow”. As the name implies, one of the propellants is injected perpendicular to the other. In our case, the hydrogen flows radially out from the inner body as the air is injected longitudinally.

As for the nozzle spike and cowl, we decided on constricting the opening at the end of the annulus to raise the static pressure after the propellants are injected, as it has the effect of increasing detonability and aiding in ignition. It ensures propellants remain within the chamber for longer, which also helps the process overall.

The Rest Of It

I would love to keep writing much more about all that this project entailed, but it ultimately was a deeply intertwined group contribution. I am highlighting my direct work on getting this RDE off the ground, so it wouldn’t be fair to any of my teammates to mention it all without also highlighting their achievements. A major part of senior design at UCF are these “Milestone Reports” where we, as a team, got together and tracked our work in a format similar to a design review. Milestone 6, the final one, contains a snapshot of the past year of our work leading up to the results we got and how we’d improve upon our work if we were to hypothetically continue testing after the semester ended. I feel it’s the best way to provide a curious reader, perhaps someone looking to base their research on our work, with the necessary in-depth info about the combustor, test stand, and our results. Please reach out to me directly if you have any questions. Email is best, all available over on the sidebar.

Here it is, weighing in at 134 pages.

To go along with the report, I’d also like to provide a quick slidestack that we used to give an overview about the project. I can’t embed it as a .pptx, so unfortunately, gifs will not work on this version. Enjoy!

Showcase Slides

However, I’d like to still cover what we experienced taking the design from a whiteboard/Solidworks to a real, working prototype. I’ll also be covering my contribution to our combined in-the-field problem solving and troubleshooting. Just know it’ll all be in a very simplified manner because, after all, we wrote that document precisely to act as a self-contained reference guide to our system. I might revisit this section and put together a quick rundown of our test stand, fluid system and DAQ and control electronics.

Testing

The testing campaign began with stress testing our control and data acquisition system by running the torch igniter over and over again. If we could pick out the bugs in our VIs, wiring, and sensors before moving onto the big leagues of trying to run the combustor with all the timings involved, we’d save a lot of pain and suffering down the road.

The torch seemed promising on its own, however, no matter what we did it struggled to resist being blown out against the main propellant flow, so at this point we decided to pivot onto the predet and continue testing the combustor itself. Look at the predet in action:

Your browser doesn't support HTML5 video. Here is a link to the video instead.

We still weren’t getting any ignition at this point, so we returned to the drawing board. After combing through the footage and data gathered so far, we were led to the conclusion that two factors were in the way. A: the hydrogen jets coming out of the inner body struggled to penetrate the column of air being injected, contributing to terrible mixing and a suboptimal local equivalence ratio. Although the design total equivalence ratio was correct and it was confirmed by sensor data, the gases weren’t being mixed due to their extreme difference in velocities. At this point we decided to increase our injector pair count from 20 to 32, as this would decrease the relative mass flow of air per injector element and possibly alleviate these issues. Look below for one of the tests from the 2nd injector iteration.

Your browser doesn't support HTML5 video. Here is a link to the video instead.

The impact this relatively simple fix is what leads me to believe that there’s still a possibility for the torch to work. I go into detail towards the end, but if the injector and mixing issues can be mitigated to the point the chamber conditions closely approximate the theoretical perfectly mixed propellant assumption, the torch would have an easier time starting a chain reaction culminating in a stable detonation. Ultimately, timelines didn’t allow me to get further testing time before graduating, but I’ll be keeping it in the backburner as I consider it novel enough to deserve further investigation and possibly a publication.

We didn’t set up back-end imaging until we confirmed we had reliable ignition, which pushed us into a time crunch to try and get the high speed camera allocated to us and set up properly. Unfortunately, we were unable to get a consistent setup stemming from some hardware issues on both the camera side of things and the mirror that allows it to look dead-on towards the exhaust. This left us with a handful of other ways to confirm operation, particularly, an increase in thrust compared to a cold-flow, and a pressure gain after ignition. Favorable data was present for both fields, but neither of these on their own could confirm whether the detonation mode was stable, how many waves were present, as well as other details about the environment within the annulus. As much as I trust our sensor data, I’m refraining from claiming we achieved detonation until we can have irrefutable evidence from studying chemoluminiscence within the combustor.

Moving on under the assumption we had a detonation mode, we varied the input pressures and timings in order to construct an operational map of the engine, telling us exactly what affected the behavior during startup and during steady state. To our surprise, pushing towards an extremely rich equivalence ratio actually improved the performance of the engine. I assume this is tied to our issues with local equivalence ratio, since more fuel being put in brings the local phi closer to the ideal range for hydrogen and air, and since the propellants aren’t fully mixed, the overall equivalence ratio matters much less in reality.

Senior Design Showcase

Beyond the milestone reports and other deliverables, senior design teams are expected to participate in the so-called Senior Design Showcase for the College of Engineering and Computer Science. Back in the day, it was a question of simply showing up to the event and talking to judges, but due to the growth of UCF’s Engineering program over the years, the showcase was split up into 2 phases. Phase 1 is the submission of a 10 minute video outlining your project goals, application, and approach to fulfill said goals. This video is then reviewed by the team’s assigned judges, and depending on the scores, it determines whether or not you progress to phase 2. If you’re selected to move on, you get to present your project at the showcase event where students, professors, and the judges get to see your display and directly hear you talk about your work of the past however many months.

I had the pleasure to lead the recording and editing of our video, as I happened to have the background in video editing necessary. I used my DVX100 camcorder to record about 3 hours of footage of our design, integration and testing as the project developed, all directly to tape and later digitized for non-linear video editing on Adobe Premiere. After wrapping up testing, my teammates and I sat down in the lab for about 8 hours combing through footage, gathering graphics and creating a rough outline to organize the flow of our thoughts. I had a great time turning this into a mind break from the cold engineering work and I hope you enjoy watching it as much as I did. It truly was the cherry on top to get to look back on our work summarized into a 10 minute video.

Final Thoughts and Acknowledgements

There’s a section covering what we, as a team, would revisit and improve with our design. I’d like to highlight some of the flaws in our work and discuss some of what I’d do to improve.

Right off the bat, we came to the conclusion that our assumptions for post-injection pressure were a gross over-estimate. This led to our annulus thickness to be waaaay below what is needed for a stable detonation wave to propagate. In a nutshell, for a given propellant mixture, post-injection pressure and equivalent ratio, there’s a certain detonation cell size. The annulus has to allow a certain number of cells to reside within it in order for the detonation wave to proliferate. This should be the easier of the fixes, as the outer body inner diameter can be enlarged in order to accomodate the new size determined by our updated models.

Moving onto the injectors, decreasing the distance between element pairs would also help with wave propagation. We started with 20 elements and during testing increased to 32. This greatly helped with ignition, but we still couldn’t achieve much more than rapid deflagration up to the cusp of a detonation mode. The spacing between the elements is still huge, and there’s not much forcing any mixing between them. If you think about the local equivalence ratio, you have a very well mixed environment directly at the injector element, degrading as you move radially towards the next element, acting much like speed bumps for the theoretical wave moving around. Slowing down and speeding up is disruptive while the combustion mode is still developing, so any reductions in inefficiencies would drastically improve the odds of detonation. This is still a fairly straight forward fix, but would require relocation of the ignition port. During the scope of senior design we did not end up stepping up in injector density, but the future work of some of my teammates as they move onto graduate research involves tackling this issue. The port can really be anywhere as long as it’s protected from the detonation front as it causes severe erosion to discontinuities in the flow.

The torch could be brought back into the mix if the conditions within the annulus can be improved, and possibly made more viable once we confirm operation of the combustor in general. However, this wasn’t the only problem with the torch. Hard starts and repeated use were very rough on the torch internals, particularly on the ceramic insulator around the spark plug. The layout currently has the plug electrode directly in the flow path of the two injectors to aid with ignition, but perhaps moving it a bit further towards the inner walls would help with thermal stresses and particularly percussive damage. Spark plugs are designed to take very high temperatures directly, but they’re fairly recessed within a cylinder and have better conduction paths to wick heat away. I also think that fine tuning ignition timing would greatly help prevent any further damage.

Now, in terms of our experimental setup, we’d be able to get a lot higher quality data by adjusting a couple things about our test stand and particularly the load transfer onto the load cells. We used S-type load cells between the thrust plate attached to the rear of the combustor and the test stand chassis. The combustor weighs in at a very generous 20 lbs, which introduces some error from gravity pulling down on it and compressing some load cells and stretching the others. We did acknowledge this in design by adding a ball transfer support at the front that allowed the RDE to rest on something while still having very low friction in the direction of thrust, although it didn’t fully mitigate off axis loading. Additionally, the propellant soft lines we used for the connection between the RDE and the rest of the plumbing were very stiff, especially when pressurized, which transferred a non-negligible portion of the thrust to the plumbing, skewing our results. My solution would be to mount the entire main chassis on linear bearings, with the load cells being relocated to the connection between this chassis and the rest of the test stand. This would allow for the internal plumbing to be negligible and for the more flexible feed hoses coming from the tank farms to be the limiting factor.

Additionally, incorporating a better loadcell calibration approach would greatly help with the thrust data issues. Currently, we used a cable and pulley attached to the rear of the thrust plate with different weights attached to map out the load cell response. Moving onto something like a hydraulic cylinder would allow for a more continous dataset to be collected to calibrate the load cells, and would be a lot easier to control than having to load and unload gym weights from our pulley system.

Overall, for a project conceived and brought to fruition in less than a year, I’m extremely proud of the end product me and my teammates created and would love to eventually apply all these corrections. I’m sure I’ll find more pain points to fix as I patch the ones I’ve just listed along with the many other listed in our report. After all, engineering is an iterative process. Solve one problem and 10 more appear.

I’d like to thank and congratulate Dr. Rezzag-Lebza, Dr. Burke, Dr. Stresau, and of course, my teammates. All our hard work paid off, and I hope we left a solid foundation for future detonation combustion research at UCF for both future senior design teams and researchers. Additionally, all the work before us that inspired the project as a whole: Fiorino, Dechert, among the many researchers across the US pioneering RDEs as a platform for future vehicles.

As a machinist, I’d also like to highlight the crucial role that UCF’s Office of Research Machine Shop played in helping us not only bring the project to life, but also playing a role as a major contributor in optimizing the manufacturability of our system. The latter, along with their hard work, allowed us to achieve a quick turnaround between iterations. We were able to go from testing and reviewing data to firing a modified engine in the span of a week. Reader, remember to thank your manufacturing team as often as you can.