Electronics: Autopilot Computer (AP)

In my last post I described the overall electronics architecture. In this post I'll give some details about one of the two Arduino-based computers: The Autopilot Computer, a.k.a. the AP.

Autopilot Computer based on an Arduino Mega
with two custom-fabricated shields

To review from the last post, I'm using two computers primarily to achieve redundancy for certain safety-related functions, namely the abilities to cut down the HAPP from the balloon and to deploy the parachutes. A side benefit is that I can dedicate one of the computers to the time-sensitive task of stabilizing the HAPP with gas jets. That's the job of the AP computer.

In a future post I'll describe the full job of the Flight Controller (FC) computer. For now simply note that the AP & FC constantly talk back and forth, checking up on mutual status. If either computer fails to answer, the other assumes a problem and terminates the flight, bringing the HAPP down safely.

The full list of hardware integrated into the AP looks like this:

• Inertial measurement unit or IMU. Detects linear accelerations and angular rates of rotation so the computer knows how the HAPP is oriented in space. Based on this information, the AP fires appropriate jets to stabilize the HAPP. I'm using a Bosch BNO055 running on the AP's I2C bus.
• Altimeter / Barometer. This is a backup for the GPS embedded into the FC computer. This altimeter is not as accurate as GPS-derived altitude, but at least we'll have approximate data stored on the AP for post-flight analysis if anything goes wrong with the FC. I'm using an MPL3115A2 running on the AP's I2C bus.
• Digital thermometer. A Maxim DS18B20. While not used for any decision-making logic on flight day, why not take the opportunity to grab some environmental data? This sensor runs on a "One Wire" bus.
• Micro SD card reader/writer. An Adafruit breakout board for logging data that can be used for post-flight analysis. Runs on the SDI bus.
• I2C hot-swappable buffer. An Analog Devices ADuM1250 running on the I2C bus. This little beauty is necessary for communicating with the FC. The two computers run off of separate power buses and do not share a common ground. In addition, the AP has several other devices on its I2C bus, and the FC would bring down the AP’s bus if it was wired in common and then lost power (try it if you don't believe me).
• Audible buzzer. A bit old school, but pretty useful for confirming certain things when the HAPP is fully assembled and we can't see the AP. It's one I had laying around from the old beginner's kit for Arduino that I bought about 18 months ago at the outset of the HAPP project.
• Bluetooth chipset. An Adafruit Bluefruit LE SPI Friend running on the SPI bus. Very useful for communicating with my iPhone during ground testing and pre-flight checks. I can send commands and receive data. Not useful once the HAPP is airborne and out of range!
• General purpose Input/Outputs for various HAPP hardware: The 12 jet control valves, the manifold pressure transducer, 4 status indicator LEDs, and pyros for the balloon cutdown (see this) and parachute deployment (see this).
• Adruino. The basis of the AP architecture. I'm using an Arduino Mega.

That's three I2C bus devices, two SPI bus devices, a One Wire bus device, 19 GPIO devices, and of course serial bus communications when the AP is plugged into my PC for programming. That Arduino is pretty busy!

I tied it all together with a collection of prototype shield boards, header pins, and connectors for some of the GPIOs.

Finally, to enable Bluetooth communications with my iPhone, I utilized once again the handy Arduino Manager from Fabrizio Boco. This little app makes it super easy to set up custom interface screens on a mobile device like this:

Yes, the "abort" button is locked out with
yellow hashes because I have fat fingers.

I'm not planning to publicly blog about the AP code development (can't give away all my beauty secrets :-) but I'm happy to answer questions via the comment section or by email. For now I'll simply refer you to some posts detailing development of the autopilot controls, located here and here. Those give an overview of the control strategy I've implemented on the AP computer.

Finally, I should note that as of a few months ago I'd never soldered a circuit board in my life. This project has given me an opportunity to gain yet another skill: Electronics prototyping. If an old mechanical engineer with large hands can learn to solder surface-mount components, then anyone can. Go grab some circuit components and have fun!

Electronics: System architecture overview

During the past few months I've been iterating furiously on designs for the electronics system. Undoubtedly this would have gone faster had I not been learning on the fly about prototyping for modern electronics. Until the past couple of months, I had never built up a circuit board or soldered surface mount components. Now I've got a dozen custom boards with different levels of quality and functionality sitting in a cardboard box (plus a few good ones installed on the HAPP!).

In this post I'll outline the HAPP's electrical system architecture. In later posts I'll fill in the details about some of the more interesting bits.

First is the design philosophy. The electronics system must fulfill the following requirements:

• Autonomy.  The HAPP must be able to conduct flight missions safely without any communication from mission control on the ground. This might entail aborting the mission if needed to ensure safety of the HAPP and of people and objects on the ground.
• Redundancy.  With regard to safety-critical systems, the HAPP should not have any single points of failure. These systems are (1) pyrotechnic cutdown from the balloon; (2) pyrotechnic deployment of parachutes; and (3) ability to dump any remaining pressurized gas for the jet stabilization system. In particular, the first item is required by FAA regulations as per e-CFR 14.2.101.35(a).
• Telemetry.  The system should provide mission control with real-time data about the HAPP's position and flight status. This is important for being able to locate and recover the HAPP after landing. It's also important for providing flight information to Air Traffic Control as per e-CFR 14.2.101.37 & .39.
• Modularity.  The system should be fairly easy to install and disassemble for testing and serviceability. 

Let's check out the overall architecture and then walk through some of the pieces individually.

HAPP Electronics Schema

The schematic is divided into five light-colored areas that roughly correspond to independent power zones, each powered by its own LiPo batteries. This may seem overly complex, but it's largely driven by the requirement for redundancy. The worst case scenario is losing control of the HAPP before cutdown from the balloon and before deployment of the parachutes. This could happen if the computer dies due to the LiPo battery running down. To protect against this possibility, the HAPP uses two computers: the Autopilot (AP) and the Flight Controller (FC), each on their own 9 volt power supply.

Furthermore, the AP and FC are in communication via their I2C busses, a common way to share information between microcontrollers. The AP and FC constantly query each other and listen for a response. If one computer stops responding, the other triggers an abort, which consists of immediately firing the cutdown pyros, firing the parachute pyros, and opening the jet valves to dump any remaining high-pressure gas. This gives full redundancy for the computers and their power supplies. The AP and FC each have some distinct duties - the AP performs the jet stabilization, while the FC handles collection of GPS data and transmission to earth via satellite link - but either computer can execute an abort.

Additionally, the FC will trigger an abort if it detects any one of a variety of problems, such as drifting out of range, staying aloft longer than expected (possible if the balloon has a small leak - it won't rise to burst altitude), or receiving an abort command from mission control on the ground.

But what if the pyros fail? To safeguard against this, the HAPP has duplicate pyros for cutdown and chute deployment for a total of 4 pyrotechnic devices. In addition, each pyro is energized by two separate solid state relays or SSRs (I’m using MOSFETS; can't energize pyros directly from data pins on the computers because the pyro igniters draw high current when they fire). Finally, the SSRs are powered from a bus that is energized by the two LiPo batteries powering the computers; if either LiPo goes down, the bus will remain energized by the other. There is no single point of failure in this setup. Failure of the cutdown or chute systems would require losing two of something: Both power sources, or two MOSFETs, or two pyro igniters.

You can also see that the cameras are separated into two power zones. The first zone consists of three GoPro Hero Session 5 cameras in parallel on a 6 volt bus (given the length of wires between battery and cameras, the cameras are happier with a 6 volt regulator instead of the nominal 5V). The second zone contains only the Garmin VIRB 360 camera. While all the cameras could theoretically run off the same 5-6V power bus, splitting them into two zones gives some redundancy: Even if one zone dies, we should be able to record something from the remaining zone. It also allows me to position the batteries independently as ballast to balance the flight dynamics. Balance has been a big issue in the past; see this old post.

In all power zones, the nominal 11.1 volts from the 3S-LiPo packs are either stepped up or down to appropriate voltages (3S refers to the number of standard 3.7 volt cells wired in series; in our case 3x 3.7V = 11.1V. Click here for a LiPo primer.) Step down conversion is accomplished with UBECs, or "universal battery elimination circuits." These are common in the hobby world of radio controlled airplanes and are readily available. "Elimination" refers to the fact that we don't need a separate battery that outputs 5V or 9V or whatever we need for a specific electronic component; we can simply use the main LiPo and branch an appropriate UBEC off of the main line. Step-up conversion is accomplished with this adjustable DC-DC converter.

As for the LiPo batteries, they are sized to allow each power zone to run for 3 hours, considering the current draw of the attached devices. Simulations indicate that the mission flight will require about 2 hours up to apogee and about 13 minutes down to chute deployment altitude, with a few extra minutes to land by parachute. By the way, the flight path simulations look like this after you throw them into Google Earth. This is approximately the view we can expect on mission day - except in 360!

With the schematic described above, the electrical system satisfies the requirements of autonomy, redundancy, and telemetry, all in accordance with FAA regulations. I believe this system architecture exceeds most hobby balloon projects in its robustness and reliability.

As for my final requirement of modularity, I'll demonstrate it with some photos of the actual system in an upcoming post.

Slight detour to Nashville

I haven't made much progress on the HAPP project during the last few weeks, and intentionally so: I decided to road trip from Michigan to Nashville and witness the total solar eclipse on August 21, 2017.

Although I wasn't able to get the HAPP in the air for the eclipse (how awesome would it have been to capture 360 video of an eclipse from high above the stratosphere!) I did manage to bring a telescope and my camera gear. The result is the following 4K time lapse video. It runs at 450X speed and compresses a 3-hour event into less than 30 seconds.

The road trip only took a few days but it took a couple of weeks to process the video. Due to thermals in the the hot summer air, the images jitter and pulsate from frame to frame. The effect was bad enough that I couldn't run automatic stabilization on the video in post processing. Instead I had the pleasure of aligning almost 800 frames of video by hand. Not for the faint of heart!

Even after hand-centering the frames, the picture still warps and magnifies unevenly. Interestingly, the image gets more stable leading up to and immediately following the totality phase. This is because the air became noticeably cooler and still as the light of the sun diminished.

Hope you enjoy the video.

Bonus:

Here are a couple of nice stills. In the second image you can see mountains on the rim of the moon behind the bright "Bailey's Beads" on the left hand side and also some solar flares peeking out on the right.

Primary camera test

Previously I posted about the 360-degree camera I plan to use for mission flights. Back then I was using a Nikon KeyMission 360, which was state of the art in 2016. Unfortunately, I was never happy with the performance, and I've been searching for a better solution ever since.

I tried a constellation of six GoPro cameras, the footage from which I stitched into a 360 video in post-production using Autopano Video software. In theory this approach offers the best possible image quality. In practice it had a couple of fatal issues. First, as small as the GoPro HERO5 Session cameras are at 74 grams each, the six cameras together were too heavy, which means the HAPP would not attain its target altitude of greater than 30Km / 100K ft. Second, the HAPP structure has too much flexibility, and when I placed the cameras all around, they vibrated out of sync. This means I could not generate a clean stitch for the 360 video.

Enter the Garmin VIRB 360. This little monster is lighter and more capable than the KeyMission 360. As of July 2017, it's the most capable prosumer 360 camera vailable. Total weight of mount and camera (including the custom-machined aluminum portion shown below) went from 340 grams with the KeyMission to 208 grams with the VIRB. The VIRB shoots at 30 frames per second, not 24 fps, giving smoother video. Furthermore, the VIRB shoots in 5.7K resolution, not 4K, resulting in a superior image, especially after applying stabilization in post-production.

Here's a computer rendering I made after drawing the VIRB in my CAD software. All components must be modeled in CAD so I can use the model for accurate simulation of flight dynamics. This is much cheaper than crashing a few prototypes. While a simple block model would be sufficient for the dynamics, why stop there? Might as well make the rendering look sweet...

My VIRB 360 arrived yesterday, and what better way to celebrate than to strap it right onto the back of my DJI Mavic Pro drone (a.k.a. the HAPP chase plane) and send it up for a little test drive? Video is a little shaky as the drone is small and vibrates quite a bit. We'll get a smoother ride on the HAPP.

Check out the 360 video on YouTube and notice the heads-up display with live flight data. You might need to open the video link in YouTube for the 360 to render correctly (click here). Also be sure to select 4K video resolution at the bottom right of your YouTube player for the best experience. Enjoy!

Upper Aeroshell in hyperspeed: Final video

Previously I posted the first two out of three videos showing how I manufacture large carbon fiber aeroshells. Here's the final video, which reveals the carbon fiber layup and vacuum infusion process. These videos run at 40X normal speed and show well over 100 hours of work in just a few minutes.

 A few final comments on the aeroshells:

• Most of the materials for the plug, mold, and part are from the terrific team at Fibreglast. Their service is fast, the quality of materials is high, and their selection is large, with almost every tool and material necessary for the entire production process.
• The final layup I used for the composite material consists of "2.5" layers. The outer layer (A surface) is this patterned honeycomb fabric chosen for its awesome looks. The inner layer (B surface) is this standard 3K 2x2 twill weave. The "half layer" consists of a few strips of this 3K twill tape.
• The resin is this special 2-part epoxy resin. I chose it due its viscosity and pot life (good for vacuum infusion), clear color (not the yellowish tint of many epoxies), and UV inhibition (the HAPP will get blasted by high-altitude sunlight).
• I don't know if you think these videos make the process look easy or not. What they don't show are all the mistakes and do-overs I needed until I landed on the final process and material selection. I made several shells that were too heavy, too thin, or incompletely infused with resin. Carbon fiber material is not cheap, and these mistakes were costly in terms of money and time. Prior to the HAPP project I had zero experience working with carbon fiber so I had to take some lessons from the School of Hard Knocks.
• Just keep in mind that I had to repeat what you see in the videos for the lower aeroshell as well!

OK, here are all three videos in order from last to first. Enjoy!


Step 3: The Part (new video posted today)

Step 2: The Mold (posted previously)

Step 1: The Plug (posted previously)

Upper aeroshell in hyperspeed: Steps 1 & 2

Here's a major post that was well over 100 hours in the making.

Way back in September of 2016, I wrote about how I fabricated the HAPP's lower aeroshell out of carbon fiber and Kevlar. There were three steps: Making the plug, the mold, and then finally the part. I also wrote about how I developed the specific layup I'm using for the composite material.

What I did not do is follow up with the story for the upper aeroshell. It took many hours to prepare the story, and for good reason - for the upper shell, I videoed every minute of the months-long process, edited it down, and sped up the playback over 40 times so you, dear reader, can enjoy it over a single cup of hot coffee without even needing a refill.

Ever wonder why carbon fiber products can be so darned expensive? Watch the videos and you'll understand! This process was not for the faint of heart.

Here are the first two: The plug and the mold. The video for the part is ready to go, but I'm out of pocket for a week and I'll have to upload it when I return. Not to mention, who doesn't enjoy a little dramatic suspense before the big finale?

This is what we're making:
One-meter diameter aeroshells.
Lower shells along the wall;
Upper shells on the floor;
HAPP structure in the middle.

The videos contain narration and they're fairly self-explanatory. The only note I'll add pertains to the elapsed time clock you'll occasionally see in the lower left corner. This is not fake. I got jiggy with Javascript expressions in Adobe After Effects, and the timer is synchronized with time codes from the raw 4K video files.

Unfortunately (?) I only filmed the fabrication work, not the design, setup, cleanup, or many of my multiple mistakes. The actual time required is about 3X what you see on the clock. I suppose if I had a small army of assistants doing those other tasks, and I already knew exactly what I was doing with zero trial and error, then theoretically I could produce an upper aeroshell in the time shown on the clock. Oh well - one can dream, no?

Enjoy the videos, and please let me know your questions and comments.

Video 1:   The plug...

Video 2:   The mold...

Onward!

Full scale prototype and system test

System testing has begun on the HAPP full scale prototype!

This flight-worthy proto is basically what will go on the main missions, although the final version will be cosmetically enhanced and will also contain a few fixes for issues discovered during testing. We're going to beat up this proto pretty good...

During the next several weeks I'll be running a series of tests with ever-increasing difficulty, culminating in low-altitude tethered flights.

First up is deployment of the Earth Landing System parachutes. You saw the ELS in action previously. This time, however, the ELS has been integrated into the main structure and we need to verify mechanical integrity - in other words, does the structure survive the recoil from the parachute mortars? In previous trials on a test bench, the recoil bent some steel mounting brackets. The HAPP doesn't contain steel; the entire structure is carbon fiber composite. Will it survive?

The production quality of this video may not be quite up to the standards of the previous one. Sorry! This one was strictly for engineering analysis. As a little bonus, this new video contains footage from the new HAPP chase plane, a.k.a. my DJI Mavic Pro drone.

Enjoy

Technical notes:

• Delay between first and last chute deployment was 50 milliseconds. Pretty good, considering that opening of the three mortar caps relies on the shearing of nylon pins. See the old ELS post for details.
• There were a few spots where the structure experienced minor cracking, all of it along epoxy-bonded joints. I will reinforce these joints with carbon fiber lapping material.
• None of the structural components received any damage. Good to go!

Structure: Crash pad

Here's a quick update on progress with the HAPP's structure.

From previous posts you might recall the HAPP has an outer aeroshell that I custom-fabricated from carbon fiber in an extremely labor-intensive process spanning multiple months. For a refresher, see my old posts about the Plug, Mold, Composites, and Part. The central structure, which includes the main strut and Earth Landing System (a.k.a. parachute mortars), drops into the top of the aeroshell and cinches into place with Velcro straps.

Between the bottom deck of the structure and the lower aeroshell is a small gap that needs to be filled with some type of impact-absorbing material. This material will help absorb kinetic energy as the HAPP touches down at landing. Which material should I use?

I played around with materials of various densities and elasticity, such as styrofoam (inelastic) and neoprene (too dense and heavy). I also experimented with expanding polyurethane foams, both the rigid and flexible varieties. I finally found the right material: FlexFoam-IT! X from Smooth-On. It's a flexible PU foam, but one of the stiffer varieties. The foam expands and hardens in a few minutes after mixing the two components together. The resulting material is spongy and fairly lightweight.

For those who care about such details, the two chemical components are a polyol (Part A) and an isocyanate (Part B). Together these react and form an open-cell polyurethane. Although Smooth-On doesn't divulge the details, I suspect one of the components contains water. The presence of H2O, after some intermediate steps, eventually results in the production of carbon dioxide gas. The CO2 acts as a blowing agent and foams the polyurethane like a mousse. FYI the giveaway for the water is the visible presence of an amber-colored polyurea "hard phase" in the material under certain conditions...one of the chemical consequences of adding H2O.

A pinch of this, a dash of that...

Next I needed a mold to contain the foam as it expands and deliver the proper shape. For the bottom side of the mold, I simply used the inside surface of the lower aeroshell. This ensured a perfectly-curved pad that fits the aeroshell precisely.

I made the walls of the mold from an old plastic bucket turned upside down. I made the the top of the mold from a piece of generic foamcore board cut to fit. Fortunately, the diameter of the bucket almost exactly matched the diameter of the main structure's lower deck, which gave a properly sized pad. I secured the foamcore board in the correct location with metal angle brackets... and the location was easy to determine now that the HAPP is fully modeled in 3D CAD.

Inserting upper side of mold into bucket.
Metal brackets are covered with foam tape.
This whole assembly will be turned
upside down during molding.

Foamcore board is a little chewed up because
I took this pic after de-molding 2 pads.

After waxing the makeshift "mold" with release agent, it was time to mix the expanding PU foam and pour it into the mold. I poured a blob right into the center of the aeroshell and quickly set the inverted bucket on top, holding it down with a heavy weight to contain the foam's expansion.

After the foam cured, I peeled it out of the mold and trimmed off any excess flash. Voilà - a perfectly sized, custom-molded crash pad of just the right density. Pad weight is 248 grams. That's a fairly substantial weight penalty to pay for good impact protection, but probably worth it. We'll see how things hold up in flight testing. It might be possible to cut some chunks out of the pad to economize on weight, perhaps leaving a thick ring with a central donut hole.

Clockwise from top left:
(1) Holding mold in pace while foam expands;
(2) Removing mold from aeroshell;
(3) De-molding the foam pad;
(4) Finished pad after trimming flash.

Below you can see how the pad is situated in the HAPP. Before inserting the structure and ELS assembly, the pad is positioned in the location where it was molded. The structure is dropped into place and the lower deck compresses the pad slightly.

Note that the aeroshell shown in the photo is the "B" version I'm using for early flight testing. I never bothered to clean up the exterior after vacuum molding the carbon fiber shell, and I've hacked it up a bit during the final stages of development. I'm saving the cosmetically perfect "A" version for a future unveiling. After all, I have to maintain some kind of dramatic suspense...

Left: Structure with ELS & 360 VR
camera, sitting next to aeroshell.
Right: Inserting structure into aeroshell.
Foam pad placement is visible.

Recovery: Cutdown pyros

What goes up must be made to come down.

On mission day the HAPP will be carried to an altitude of 30 Km / 100,000 ft by a weather balloon. Once it reaches apogee, we need to cut the umbilical cord between the HAPP and the balloon. There are two possible scenarios.

First, the balloon may burst as it expands in the near-vacuum conditions, in which case we need to cut the cord and release the floppy balloon to ensure a stable descent and a clean deployment of the parachutes. This is the most likely scenario, and in fact is our Plan A for returning the HAPP to Earth.

If the balloon does not burst, we need to cut the cord to ensure the HAPP comes down and doesn't turn into a perpetual "floater" that proceeds to circumnavigate the globe until ultraviolet radiation finally degrades the latex skin of the balloon. In fact, FAA regulations require a redundant Plan B for unmanned balloons for exactly this reason (Federal Code of Regulations, Title 14, Part 101, Subpart D, 101.33).

Most folks I've seen on the internet use a NiChrome cord cutter. This type of system takes a small piece of NiChrome wire, spirals it around the cord (usually made of Nylon), and at the moment of truth sends a relatively high current through the wire, thereby heating it up and melting the cord. Here is one commercially available system, and here is a video of someone's hobby version in action.

As the 9 kg HAPP is heavier than most hobby balloon payloads, we need to use a fairly robust umbilical cord: Kevlar instead of Nylon. However, Kevlar has a melting temperature of about 500C/930F, whereas Nylon 6 melts at 220C/428F. Thus, we'd need to get the NiChrome almost twice as hot to melt the Kevlar, which means more current, bigger batteries, more weight, and lower reliability.

Here is where we have another opportunity to use pyros and blow stuff up (and those are always welcome opportunities :-)

After exploring a few alternatives for home-made, explosively actuated guillotines and the like, I decided to keep it simple and use the lovely little Cable Cutter from Prairie Twister Rocketry. This device is designed to cut plastic zip ties, but I figured it might work for Kevlar string as well.

You can glean the operating principle from the picture below. Moving left to right, we have 500-lb yellow Kevlar braided string threaded through two small radial holes in a blue anodized aluminum cylinder. A small steel fire pin goes into the cylinder. Next comes about 0.1 cc of black powder (not shown, but it goes where my finger is located in the picture). After that we insert a J-Tek electronic igniter which has been threaded through a black steel end cap. To complete the assembly we simply tighten the endcap onto the cylinder. The entire assembly weighs only 10 grams. The igniter is triggered simply by hitting it with 5V from the flight control system.

You might wonder whether the powder charge will burn effectively in a nearly zero-oxygen environment at flight apogee... it's a reasonable question. The answer is that the potassium nitrate (KNO3) in the black powder provides the oxygen needed for combustion, so the powder is essentially self-oxygenating. Guns can be fired underwater for the same reason.

Cable Cutter from Prairie Twister Rocketry

So does it successfully cut the Kevlar cord? Check out the video below.

Looks like a solid solution. We'll use two of them on flight day for redundancy.

FYI, a rough analysis of the video shows the cable cutter jetting away at about 90 kph / 55  mph  as the hot combustion gas shoots out of the endcap...

Onward!

Recovery: ELS / Parachutes

First the goodies. Check out this video to see the first system-level tests of the HAPP's parachute deployment system. Video analysis shows those chutes firing out at about 120 kph (75 mph). The video is a bit silly - gotta have fun while you work - but read on down below for the serious technical details.

The project is moving right along toward a summer mission flight. Flight controls are done. Structure is fully designed, parts have been machined, and final assembly is underway (details to follow soon). The next major system to validate is what the Apollo Program folks called the Earth Landing System or ELS, a.k.a. the parachutes.

From the rendering below you can get an idea of how this is supposed to work. Protruding from the upper deck of the HAPP are three mortar tubes which contain the parachutes. They're nestled between the vertical strakes or fins. During the mission, the HAPP will blow the chutes at a low altitude after free-falling from the edge of space. Most "space balloonists" use a chute that's always deployed. While that approach is simpler and less risky, those crafts can drift a large distance away from the launch site as they descend slowly through 30 Km of windy atmosphere while trailing a parachute.

Overall HAPP rendering

ELS shown in isolation

We're going for rapid descent. While the risks may seem high, consider the math. The HAPP weights only 9 kg and has an aeroshell that's 1 meter in diameter. In essence, the HAPP is a round wing. I calculate terminal velocity without chutes to be around 47 kph, or 30 mph for the Americans. If all three chutes fail, the craft will suffer damage on impact, but the data payload (4K, 360-degree virtual reality video on a micro SD card!) will most likely survive, and the HAPP is not going to do major damage to buildings, cars, or people.

If at least one of the chutes deploys from its mortar, the ground impact speed will be even lower. It doesn't even have to inflate - a tangled chute makes a fairly effective streamer. If all three chutes inflate successfully, the impact speed will be the same as dropping an object from about 30 cm in height (about 1 foot). Pretty cushy.

The chutes I'm using are Iris 60" Ultralight chutes from Gene Engelgau and the good folks at Fruity Chutes.

As for the deployment mechanism, I spent several weeks toying around with various concepts. I tried springs, elastic bands, and even a simple method that relied on the rushing wind surrounding the HAPP to pull the chutes out of their containers, assisted by small drogue chutes. None of those methods worked well. They were either too heavy, or did not eject the chutes with enough force, or both. The chutes need to be thrown forcefully clear of the HAPP so the canopies have time to inflate before the wind carries them up and tangles all the lines together.

In the end I decided to get with the times and do a proper mashup of various technologies. For the deployment force, I adopted the approach of Cameron Tinder over at Tinder Rocketry. He has a system called the Raptor that was designed for model rockets. It uses redundant electronic igniters to light a black powder charge which in turn drives a sharp metal spike into a canister of compressed CO2. Once ruptured, the canister vents the CO2 into the parachute compartment and eventually forces out the parachute.

For the packaging, I went with the approach of the drone recovery system available here from Skycat. Skycat uses carbon fiber tubes to hold the parachutes. Chutes are deployed using a proprietary piston system, which is interesting, but the applicable model from Skycat is listed at 0.32 kg without a parachute. To get redundant systems and equip them with chutes would result in a heavier system than my approach. Plus, where's the fun in buying a finished system when you can waste weeks building your own?!

However, unlike the Raptor's single chamber, I need to direct the gas into three separate mortars, so I designed a new manifold for the Raptor and had the awesome Jason Davis at Pioneer Cuts machine the manifold on his CNC mill. In the photo below, the manifold is the silver aluminum part into which the CO2 canister screws. It directs the gas to the mortars using high-pressure red macro lines and push-connects. My manifold mates to the charge and fire pin portion of a standard Raptor. You can see these standard parts in the photo below: The red anodized aluminum part, the spring, and the small cylindrical steel parts in between.

A tech mashup: Part Raptor, part Skycat,
part home-brew, and 100% fun.

Let's walk through the details.

When it's time to blow the chutes, the flight computer energizes the igniters. The igniters set off the black powder, which in turn drives a steel fire pin into the CO2 canister. Gas vents out into the manifold and gets routed to the mortar tubes by the high pressure macro lines.

As the mortars pressurize with gas, the caps are pushed outward with significant force. Each of the caps are secured with four #2-56 nylon screws. These plastic screws are tiny, and eventually they shear off as the pressure accumulates. This allows the caps to fly off and the parachutes follow behind them. Getting all the math right was a fun exercise - what size of CO2 canister to use, what size of nylon pins, diameter and volume of the mortar tubes, etc.

For those who care, the theoretical maximum force required to shear four nylon pins is 603 N, which is equivalent to a 61 kg / 135 lb weight pressing on each of those three little mortar caps. The 45-gram CO2 cartridge I'm using, a Leland 87202, has enough gas to generate 3382 N on each cap should the pins prove stubborn or the caps get jammed, perhaps if the sealing tape around the rims freezes up at high altitudes. That's a safety factor of 5.6. I was shooting for 5, but cartridges only come in certain standard sizes.

I also put some one-way check valves in the macro lines. They're the black, blue, and silver cylinders located midway down the red macro lines. These valves allow gas into the mortar tubes but they don't let it flow out. This is critical, as the nylon pins are each a bit different (ah, the beauty of polymer chemistry), and they will shear at different times: In the video you can clearly see delays up to approximately 10 milliseconds. If one mortar opens first and there are no check valves, gas will leave the unopened mortars, flow back through the macro lines, and rush out of the open mortar. The unopened mortars will fail to deploy.

The mortar tubes are 9" sections of Size 18 INFINITubeV from Rockwest Composites. The caps utilize 1" sections of Size 17 tubes. The end plugs are machined circular sections of 2.2mm thickness carbon fiber plating. Plugs are bonded to the tube sections with this specialty 2K epoxy.

All in all, the total system weight including parachutes came in at 1006 grams. That's a little over 2 pounds for 5.4 square meters / 59 sq ft of total parachute canopy, explosively discharged, with redundancy for the chutes, the mortars, and the black powder igniters. Potential single points of failure are the CO2 canister and Raptor fire pin, but that's a risk I'm going to take.

Here's a picture of the chutes ready for folding and stowing in the mortar tubes.

Foosball table is optional for
successful parachute deployment

So that's about it for the Earth Landing System. Onward!

Structure: Carbon fiber cutting and bonding methods

Ever since I finally got the internal structure re-configured for optimal balance and aerodynamic stability using some serious 3D modeling, I have been working frantically like Doc Brown jacked on Red Bull to get the structural components built and assembled.

Soon I'll post a walk-through of the entire structure and the various components, but for now I thought I'd show you what the parts look like and discuss how I processed them. This picture shows the whole suite of structural components, which are all fabricated from carbon fiber.

Need more fiber in your diet? Carbon fiber, that is.

The closed sections (round and square tubes) were purchased from Rockwest Composites and cut to length. The round tubes are mortars for the parachutes, and in the photo you can see one of the 60-inch parachutes peeking out of its mortar tube - stay tuned for excellent videos of explosive parachute deployment! The square tube is the main strut that runs vertically down the center of the HAPP.

The flat sections, also from Rockwest, were a bit more tricky. These are, variously, the internal decks (round), three aerodynamic strakes (triangular), and holders for the parachute mortars (various small pieces). I purchased rectangular plate stock of various thickness from Rockwest and had to machine the parts using a CNC mill. Thickness ranges from 0.7 to 2.9 mm, depending on the part. The largest single plate was 24x24 inches and 2.2 mm thick (SKU # 403-22). Not cheap!

Although I've received my checkout to run the large CNC flatbed router at Maker Works in Ann Arbor, Michigan, in the interest of time I had the awesome Jason Davis at Pioneer Cuts do most of the work. This is what the process looks like:

To save weight and maximize flight altitude, I had to abandon the aluminum mounting system described in a previous post. Now I'm bonding the structure together with specialty adhesive - in particular, this 2K epoxy from 3M. This approach means the craft will not be easy to repair if any parts are damaged in flight testing, but such are the engineering tradeoffs in aerospace!

It also means the assembly process is a bit more complex. To hold the parts together while the epoxy does its bonding magic, I decided to 3D-print some assembly jigs using high-accuracy SLA. Here's one of the jigs. This one holds the end caps onto the parachute mortar tubes. I made a total of 5 different jigs. Thanks as always to Steve and the great 3D printing team at ThingSmiths in Ann Arbor.

3D printed assembly jig on
end of parachute mortar tube

All together, what you see laid out on the table in the first photo is the entire structure of the HAPP. It weighs in at 1671 grams - not bad for a craft that's 1 meter in diameter, almost as tall, and will support a lot of serious hardware, including twin pressurized gas tanks, pneumatic valves, cameras, a custom-molded carbon/kevlar aeroshell, and a few other important bits (like the pyrotechnic-deployed parachutes!). All of which will be exceeding Mach 1 on a plunging descent through the atmosphere after flying to the edge of space...

Passing 35,000

Back in May 2016, I wrote a post describing the geographical composition of my audience as I crossed 10,000 readers. I had every intention of refreshing the data periodically, perhaps as I hit 25,000. Well, I should have checked the stats sooner, as not only did I blow by 25,000, but I should hit 40,000 in a few weeks!

The total today stands at 37,419. As I did back in May, allow me to give some color, followed by some awards to you, my dear readers...

For starters, here are some basic stats:

• Top posts: (1) Valves and pneumatics, (2) Controls finale
• Top referring site: (1) Facebook, (2) Google
• Top browser: (1) Chrome, (2) Safari
• Top OS: (1) Mac, (2) Android

And here is the geographical breakdown:

Back in May I pointed out the curious (to me, anyway) fact that the highest source of readers outside the USA was Pakistan. Mystery solved: In the description of the blog for Google's search engine, I had included the keyword "bang bang." This is a reference to the particular type of control algorithm I embedded into the flight computer's autopilot function, explained here. It was an innocent but fateful mistake, as folks in Pakistan apparently search for popular Bangladeshi celebrities using the word "bang" followed by a name, and for whatever twisted reason harbored by its nefarious AI, Google was returning a link to the HAPP Blog amidst the search results. I know this to be true because Google shows me the search terms used by those who clicked through to the HAPP blog website!

For those curious, the #1 most searched-for Bangladeshi celebrity by those who also visited the HAPP blog is this lovely person:

Apu Biswas, Bangladeshi actress

Alas, with the "bang" reference deleted, the Pakistanis have fallen out of the Top 5 (although someone apparently has enough spare time to continue visiting the blog). The crown now belongs to France! Here's a special tribute to mark the occasion:

Vive la France!

Finally, I simply must give an honorable mention to a recent newcomer on the list. I have ZERO idea how folks in Oman found this blog, but my dear Omani readers - I salute you!

Cap'n, we need more power!

Now that I've started testing the field of view and placement of cameras - including the primary camera, the stonkin' Nikon KeyMission 360 - I've run right smack into an unfortunate reality: My old PC is useless for editing 4K video. It literally cannot run the editing software without immediately crashing. As the primary goal of the HAPP project is to get 360-degree 4K video and give my dear readers a virtual reality ride to space, this is a pretty big problem.

In addition, I'm having difficulties manipulating the full 3D model of the HAPP in AutoCAD Fusion 360. It crashes frequently and I've lost hours of work. For that matter, the flight control simulations in MATLAB/Simulink are painfully slow.

As I have lots of work remaining in all those areas, it's time to bite the bullet. I'm retiring the PC I built nearly 7 years ago (it was a banger back then!) and building a new rig this week that can handle these tasks.

For those curious about such things, here are the parts I just ordered today:

An 8-core CPU, a screaming graphics card, 64 GB of RAM, 12 TB of storage, and lots of liquid cooling ought to hold me for a while :-)

It sucks to deviate from the parachute-testing plans I had for the coming weekend, but this is a necessary step...

Tether and camera mount: Initial design

In the last post I described how I learned a hard lesson about taking time to do a thorough engineering analysis when necessary. In this post I'll apply that lesson to a small but critical part of the HAPP: The tether ring, which is where the HAPP is attached to the balloon that will lift it over 30Km into the sky. It turns out that this part is one of the most highly engineered components in the entire project.

From the start of the project I knew I'd need a robust place to attach the balloon. Initially I assumed this would be a simple metal ring of some sort. However, after completing the full 3D model of the HAPP, I decided to integrate two additional functions into the tether.

First, the tether will also serve as the attachment point for the parachutes that will deploy during descent. There's no sense adding additional mass for a completely separate structural element to achieve this function.

Second, the tether will also contain the mount for the primary camera, which is a 360-degree 4K ultra HD camera (specifically, the Nikon KeyMission 360). The only place where the camera will have a mostly unobstructed view of the world is from the top of the HAPP. If I attach the camera to, say, the side of the main strut, then a large portion of the view will be occluded. The camera therefore needs to go at the apex, but it can't interfere with the tether or the parachutes.

After several iterations using 3D printing to verify the geometry (thank you Owen and Steve at ThingSmiths!), I arrived at the following design:

In this rendering you can see a few important features. First, note that the base of the part will insert directly in to the square-section carbon fiber tube I'm using for the main body strut. The rendering shows four of the M3.5 mounting bolts that will go through the strut wall. The bolts are mainly to ensure position while I bond the metal body of the tether to the carbon fiber strut using some specialty adhesive.

One side of the camera - and one of its lenses - is visible in the rendering. This camera is symmetrical and has a second 180-degree fisheye lens on the other side. The camera automatically stitches together the two 180-degree videos to form a 360-degree virtual reality video in all its glory.

You can also see two of the three parachute attachment rings at the ends of short arms that jut out horizontally. There is the main ring enclosing the camera, and finally there is the actual balloon tether ring sitting on top. It's not obvious from the picture, but the parachute rings and camera enclosure are designed to be almost invisible to the camera's field of view. Here's a quick demonstration. In this 360 video, note the kevlar/carbon fiber lower aero shield that I spent the entire summer of 2016 learning to fabricate...

360 camera field of view confirmation.
If this doesn't play on your screen in 360,
try viewing with YouTube here.

What you can't see is how the camera is attached. There's a 1/4"-20 UNC bolt going up into the camera's standard mounting socket. However, as the part will be bonded to the main strut, the bolt must be started and tightened using a long tool inserted from the bottom of the 836 millimeter-long main strut. To facilitate this blind insertion, I added a "docking cone" to the bottom of the part, which is visible below:

Bottom view of tether / camera assembly
showing blind insertion cone for 1/4" bolt.

With the basic geometry defined, it was time to optimize the part for light weight while insuring it can withstand the forces resulting from opening shock of the parachutes. A skinny part is better as it minimizes total weight and maximizes ultimate flight altitude, but more importantly, this part is at the apex of the HAPP and will strongly contribute to a higher center of gravity. This is bad - see my struggles with CG in the last post. Too skinny, however, and the part will fail when the chutes deploy. How to find the right balance?

For this we need some real engineering, and the right tool is finite element analysis. The 3D modeling package I'm using, AutoCAD Fusion 360, happens to have a nice little FEA component that's extremely intuitive and easy to use. To run the analysis, however, we need to know what maximum force we can expect from the parachute opening shock.

I performed the shock analysis by hand using some references from NASA back in the 1960s (the technical gift that keeps on giving!), the Naval Ordnance Lab, and the Naval Surface Weapons Center. I used numerical integration to scale time/force curves given known parameters, yielding a conservative maximum force estimate of 1150 newtons.

If you want to play around with such things and don't feel like rolling your own, I recommend you check out a little piece of software called OSCALC. It gave a worst-case force of 1050 newtons, thereby validating the accuracy and conservative nature of my approach. It also comes with a user manual that walks through some of the math.

The analysis assumes we are using three of these lovely parachutes from the folks at Fruity Chutes. The number and size of chutes was selected to yield a landing speed of 2 meters per second, which is equivalent to being dropped from a height of 0.2 meters. If one chute fails to open, the landing speed will be only 3 m/s, which is still pretty cushy.

With the forces defined, it was time to optimize the design using FEA. I started with a design based on manual calculations using cantilever beam equations. From there, I went through 7 cycles of fine-tuning the design and the choice of materials. Finally I settled on aluminum 7075-T6 and the design shown in the renderings above.

Here's a little video showing the final design after optimization. The amount of deformation is greatly exaggerated - in real life you'd not be able to see it with your eyes. The blue color denotes a high safety factor, and yellow, orange, and red colors denote progressively lower ones. A safety factor of zero means the part is right at the yield strength for the material. I optimized the design so the region with the lowest safety factor is at 0.5 - appropriate given the conservative analysis of parachute forces.

Analysis of safety factors using finite element analysis.
The short video show how the arms deform as
the parachutes deploy.

Same analysis, just a snapshot instead of the video

To give you an idea of how the design evolved with the FEA analysis, here's a comparison of the parachute attachment arms from version 1 and the final version 7. The arms got shorter, they changed from a round profile to a variable elliptical profile, and the chute attachment rings got substantially smaller. All of that extra metal was not necessary. Besides changes to the arms, the base portion that inserts into the main body strut got longer. This was mainly to accommodate the cone-shaped feature that assists with blind insertion of the camera mounting bolt.

Evolution of parachute arm
from v1 to v7 using FEA

Final part weight is 139 grams. Currently, the part is being CNC milled out of a single piece of 7075-T6 aluminum by the outstanding team at ProtoLabs.

OK kiddies, this concludes our case study in mechanical engineering for the day. See you next post for some live testing of parachutes!