eVTOL

Introduction

Creating an eight-motor electric aircraft that a human could ride in is hard. But keeping its empty weight under 115 kg is something to be reckoned with. What is more representative of the future than flying cars? Well, eVTOLs are the closest we have come so far to the dream being realized. Start-ups and individuals alike have been taking advantage of FAA Part 103, which allows qualifying ultralight aircraft to be flown without a pilot's license. If you could make a 115 kg aircraft with a max speed of 55 knots, a power-off stall speed no greater than 24 knots, and a fuel capacity of no more than 5 U.S. gallons, then you could take to the skies.

Buying one is possible. Several companies in America and abroad sell them. In America, you could fetch a Jetson ONE for $148,000 and sit on a multi-year waiting list. Or you could get one for half the price from Alibaba, but that may be scarier than making it yourself. If you were to make it yourself, it would be expensive and dangerous, but it might also be cheaper than the Jetson ONE and more trustworthy than sketchy marketplace alternatives with limited documentation.

I attempted to design one myself at lower cost while accounting for the propulsion, battery pack, structure, controls, and safety systems required to put a person in the air. Looking at who has come before, many had short battery lives and a bit of risk involved with the open propellers. I wanted to see if I could build one of my own, and set out to determine the feasibility of such an aircraft for my own backyard.

Propulsion

When selecting the motors, propellers, and ESCs for my design, I largely considered the cost, physical footprint, rated thrusts for continuous hover and max power, and general power consumption.

The main candidate motors were the V13L, V10L, and U15 series motors by T-Motor and the Hobbywing H13 coaxial motors. While there were appealing aspects to each of them, the V10L and U15 motors were simply not powerful enough for a 215 kg aircraft. Choosing between the V13L and H13 meant weighing major tradeoffs. For instance, while they had roughly equivalent capabilities in regard to continuous hover, they differed greatly in cost and physical footprint. The differences in max thrust were only considered for what the relative thrust was for continuous hover. So if hover is 50% of max thrust versus 70% of max thrust, 50% is considered better for various heat and control reasons talked about in the next few paragraphs.

Between the H13 and V13L, the H13 is designed for agricultural drones so it had a much larger footprint, requiring 52 inch propellers. However, it consumed less power, was significantly cheaper, and had the ESC and both coaxial motors baked in as one product, so that it was very easy to plug and play into the system as a whole. The V13L was much more expensive, required a dedicated mounting system, and required buying the ESCs and the two V13Ls separately and putting them together. However, the V13L was significantly smaller at 42 inches, which brought the footprint of the vehicle as a whole down nearly 2 feet. Plus, when considering each arm as a whole, including the propellers, the cost of the two systems came out closer than originally thought. After some time I ultimately decided to go with the V13L, with the ESC and propellers that T-Motor recommends for it, the Thunder 12/24S ESCs and VZ42x16.5 props.

These represented a significant part of the projected costs, being around $18,000 for the 8 V13L motors, Thunder 12/24S ESCs, and VZ42x16.5 props, with a collective mass of 20.5 kg of the total 115 kg budget.

These motors also determine the kinds of forces the arms will be required to endure, around 110 kgf of max force on each arm, and the size of battery, 100 V and around 50 to 60 A per motor. Some design considerations I also looked at with the motors is that I wanted it to be hovering around 50% of max throttle so that the motors don't burn out, are less subject to wear, and have more upward and downward thrust margin for stabilization. In addition to that, there needs to be some room for voltage sag and the ability to still be able to lift the aircraft without overheating. This is because when voltage drops, amps have to rise in order to match that same power output, which increases the heat by I^2R and can push the ESC and batteries closer to limits quicker. Testing the motors to see how they hold up to the manufacturer test data would provide more accurate thrust and power assessments, but until then I will be trusting the manufacturer data.

Further, due to the coaxial design, there is some efficiency penalty from the lower motor and prop pair being in the disturbed airflow from the top. This penalty wasn't calculated, but I acknowledged that it could be a later consideration, and I chose a motor that would allow prototyping in that regard so that the lower motors would be operating at a higher power for equivalent thrust relative to the upper motors without overheating or reaching their limit. Another reason I preferred the smaller propellers was so that the aircraft could be better controlled and be more responsive in comparison to the 52 inch propellers, which may have made it feel more sluggish while providing more g/W efficiency.

Battery

The battery requirements come directly from the motor selection and an arbitrary goal I set for the aircraft, to be able to fly for 15 minutes. The motors require 100 V and roughly 50 A each. The 100 V translates to a 24S battery, which gives 100.8 V full charge and a nominal charge of 86.4 V. I am choosing lithium-ion class batteries for energy density and manned endurance. However, for amps, the issue of how to split up the parallel batteries came to the forefront of testing.

I considered various battery architectures, such as an all-one battery system, 8 batteries in series and parallel around the aircraft, and two systems with 4 batteries for the top motors and 4 for the bottom motors in the coaxial configuration.

By all-one battery system, I mean similar to an electric car, where the battery would be contained along the bottom of the aircraft and have all the motors and high current ESCs routed from there. This would be simpler and a more streamlined manufacturing process, since it is one part. The major drawback is redundancy and cooling. If it were all concentrated right under you, there would be 400 to 500 amps continuously flowing out with no good airflow solution, since you would be sitting on top of it.

Then I considered spreading 8 battery modules across the aircraft so that there might be more redundancy in the system if one of them broke, albeit only on the ground, and so you could replace it more easily and charge them individually. Further, there could be better heat distribution and cooling solutions with them spread out. Spreading them out also distributes the center of mass better, as these will be nearly 50% of the weight budget.

Finally, I considered what would happen if the battery suddenly broke midair, and considered the size of wires you would need to transfer 400 A. To help both of those issues, splitting the system in two could help in both areas. With battery system A supporting the top motors and system B supporting the bottom, if one fails you could at least run the other at a much higher power, and you could look into control systems that allow a semi-controlled descent, even if that means hurting the battery or the motors. That would also mean there is only 200 A going through each system. While still a lot, it is more manageable. Some consideration was given to whether you should go further with this and split them up even more, but I figured that could be tested and simulated more effectively later in the process.

With the very general architecture decided, I looked at various battery cells and pack options for my VTOL. Among the top choices were the Molicel P42A, the Samsung 30Q, the Molicel 45B, the Samsung 50S, the Molicel P50B, and then some other full packs such as resold Tesla Model S packs, Alibaba options, and the Tattu Pro LiPo. They all give roughly 3.6 V, so they all need the 24S battery layout. Then I looked at energy density, cost, and different layouts. With the prebuilt Tesla packs, the battery wasn't optimized for flight and was too heavy for our purposes. The Tattu Pro LiPo was in consideration for a similar ease of manufacturing, but the LiPo battery type made it unappealing for long-term use. The various battery cell options meant that I would make and design the battery pack myself.

Comparing those, I came to the conclusion that the Molicel P42A had the best combination of energy density, cost, and current per cell. The Molicel 50B and 45B may be better in a lot of regards, but they were harder to obtain and significantly more expensive. The Samsung 30Q and 50S were attractive because they were very energy dense, but required many more cells due to their low current capability compared to the P42A style cells.

Set on using the Molicel P42A, I tested various 24SXp options. With X being the number of parallel cells, I tried 30p, 35p, 40p, and 46p battery pack configurations. The 30p configuration was the best on weight and cost. The weight was 48 kg for about $1,980 in cell cost alone, a nominal energy of 10.9 kWh, and a theoretical max current of 1350 A. However, nominal energy assumes you're taking the battery to 0%, which is probably not wise, so let's build in a usable capacity of 80%, with the rest treated as empty. So usable energy is 8.7 kWh. Since hovering won't be constant and the VTOL will be climbing, descending, changing speed, adjusting, and so on, the required power of the motors will be changing. With that knowledge, I estimated flight times at 40 kW, 45 kW, and 50 kW of about 13.1 minutes, 11.6 minutes, and 10.5 minutes respectively.

Taking this into account with the architecture decisions made earlier, this would be split into two systems of 24S15p to power the upper and the lower motors, each split into 4 6S15p modules. Not quite achieving the 15 minute flight time goal, but with weight and cost considerations that is alright. The cells alone would weigh 48.24 kg, which is 42% of the 115 kg weight budget. That is not counting the busbars, BMS, wires, insulation, and all the other electrical components connecting and making this battery pack, which could push it closer to 50 to 60% of the 115 kg weight budget.

With that last part, I realized that it would be hard to manufacture this battery for a first pass design. So I decided to look more into the Alibaba options, as they were cheaper and enough for a prototype. While the above architecture with the Molicel P42A could be a future project, a first design would be better off using a premade battery or set of 8 batteries, so that more effort can be put into making everything else work and removing variables that can fail. This is also where the weight budget starts to look tight. I add all of it up in the conclusion.

Chassis

While these were going on in parallel, I was designing the chassis in SolidWorks. The chassis is developed similarly to a racecar spaceframe, in that it has to be strong and stable enough to contain and distribute all the forces from the motors without bending or destroying itself, while keeping the pilot safe. It is a welded frame made from tubular aluminum 7020-T6, drawing inspiration from the Jetson ONE aircraft. It would have to distribute roughly 60 kgf from the two motors on each arm during hover, and 110 kgf at max thrust, through the primary members of the frame. I decided to use SolidWorks weldments to design this because of the rapid prototyping ability within it and prior knowledge of how to use it.

Aluminum 7020-T6 was used because this has to be ultralight for the Part 103 requirement, and since the battery and motors are already taking up so much of the weight budget. For the primary chassis tubes, 1.5 inch diameter tubes with 0.063 inch thick walls were used, as it seemed through FEA that diameter increases were needed more than thickness, since tube bending stiffness scales with radius more than thickness. However, depending on the crushing safety and requirements of welding the members together, there were versions made with thicker tubing, as much as 1/8 of an inch.

The chassis was designed with a high factor of safety as an aim, especially in the prototyping phase of the aircraft. The design relies heavily on triangulation, shear transfer, and load paths through primary elements to ensure a secure ride in the eVTOL. This was the guiding force for the arm trusses. The arms would connect at a node cluster and distribute the force across the aircraft through the upper and lower rails, diagonals, and side triangulation. This allowed for better spaceframe behavior and ensured the arm wouldn't be dependent on one mount. The arm would behave largely as a cantilever, so a rectangular 2 inch by 3 inch tubing was chosen for stiffness benefits.

The arms would be about 1 meter long, with the propellers adding additional size, so hinges were considered to allow the arms to close and make it easier to transport the VTOL. The most promising design was one with inserts that would be secured into the arm tubes, and then a closing and locking mechanism with full positive stops that would allow most of the force to travel along the desired load path while minimizing force into the hinge part. The hinge is not intended to carry the main flight bending moment. The positive stop and receiver geometry is.

After a few versions and iterations of the model and parts, I decided to assemble a 1:5 scale version out of wooden dowels and glue. This was to see ease of manufacturing with similarly round and rectangular pieces, while replicating how welding geometry and sequencing would work with glue substituted.

While it went mostly as planned, the major realization was the nonstandard cut lengths that came from using the trim and extend features in SolidWorks weldments. More care would be required to ensure that the lengths were rationalized to easier numbers, and that they were standardized so that two members with near equal unique member lengths could simply be the same length for further ease of manufacturing. The model largely validated the geometry and helped me understand the assembly sequence of the parts, along with some ideas about how to fix the more awkward joints.

Safety Issues

In addition to this, I tested some drone software and put together a regular drone to see how it could be extrapolated to larger versions. Trying Betaflight and ArduPilot, I found ArduPilot was better matched for this safety-first and autonomous prototype of the VTOL, and safer for if a person were to ride it. This drone also made me realize that there would need to be serious safety efforts and better methods for testing and calibrating a full sized VTOL, so that it could be made safe before anyone got in it, and to ensure everyone was safe while calibrating it.

Even done perfectly, the safety issues with VTOLs in their current state seem large. Having four one-meter-long coaxial propeller pairs around you is a serious safety risk no matter how careful you are. Accidents happen, and crashing with one of those around is a significant obstacle to overcome, along with the possibility of rotor or battery failure. There may be a way, but it will likely take a different form from drone-style flight.

Conclusion and Next Steps

Adding up what I sized, the design does not close. The motors, ESCs, and propellers come to 20.5 kg. The battery cells come to 48.24 kg. The best version of the chassis was about 20 kg. That is 88.7 kg of the 115 kg budget from three subsystems, and the battery figure is only the cells. Using my own estimate that the pack reaches 50 to 60% of the budget once the busbars, BMS, wiring, and insulation are counted, the total lands somewhere between 98 and 110 kg. What is left for the seat, the flooring, the restraints, the controls, the avionics, the wiring runs, the ESC mounts, the arm hardware, and the landing gear is 5 to 17 kg. That is not enough. Getting under 115 kg would mean a denser cell chemistry, a smaller pack and a shorter flight time, or giving up on Part 103 and building to a different set of rules.

There were also limitations with the spreadsheet calculations done to evaluate batteries, energy, and heat output. A more robust solution would be to use MATLAB Simscape and Simulink to more accurately simulate the VTOL as a whole and systematically test different options to arrive at proper parameters. The motors, batteries, and thermal modeling could work as a whole through this instead of as estimates. That would also settle the weight question with something better than a spreadsheet.

No prototype got welded due to the mounting costs the project would require and the general safety issues. While this was unfortunate, it served as a strong feasibility study on how far these kinds of devices have come, and it is reasonable to hope that as materials and batteries continue to improve, a serious effort to bring about new modes of transport, and hopefully traveling as the bird flies, will become more realistic for the future.