HVAA Pod

→ mechanical design, test setup development, thermal qualification, supplier sourcing 🛠️

→ jul 2022 - jun 2025 📅

role

I was a mechanical engineer and later project engineer on the HVAA Pod. I owned three major workstreams: building a standardized test setup to quantify the performance of a ram air turbine a senior engineer had designed, maturing an unfinished clutch design from concept to a manufacturable, qualified assembly, and the full mechanical design of the pod’s controller assembly, from envelope allocation through ICD handoff to the PCB layout team.

turbine test setup

The turbine had been designed but never quantified against a real standard. I reverse engineered the AMCA test setup used for fans and pumps and retooled it to qualify a turbine instead. The setup measured air pressure and temperature using proven methods for calculating airflow and power, and monitored both the airflow entering the turbine and the power it extracted. I built a small wind tunnel: a large fan pushed air through a 10 foot tube into the ram air turbine, which spun and generated torque that I measured with a rotary dynamometer mounted to the back of the turbine. Ground-generated airflow doesn’t match what the turbine sees in flight, and the dynamometer had a maximum power input limit, so I upscaled the estimated airflow requirement to account for both.

clutch design: spring sourcing

The clutch had a basic design premise from a senior engineer, but much of it wasn’t manufacturable. The steel spring in the original design was a fully custom spiral wound spring (3" diameter, 1/4" wide, ~30 thou thick) quoted at $2,000 each, which was a non-starter. I went through hundreds of candidate replacement designs, evaluating load capacity and corrosion resistance, before finding a niche rotary spring supplier in Sweden whose COTS spring hit every requirement: exact torque, exact turn radius, and our form factor. The supplier wouldn’t respond to any of my messages, so I found everyone affiliated with the company on LinkedIn, tracked down the president, and emailed the president and CEO of the distributor directly. That got the sales team responding. I then worked with internal supply chain to get approval to buy the foreign COTS spring. Final cost: about $5 per spring, down from $2,000.

clutch design: thermal qualification

My second task was to thermally test the clutch the previous engineer had selected. I built a test mechanism to verify the clutch could retain a target torque (100 in-lb) across the full temperature range, from -55°C to 55°C. At -55°C, I immediately found icing issues. The clutch also came back from -55°C corroded once returned to ambient — the wrong material had been selected for the application. I ran a trade study on finishes and landed on a custom finish for the clutch, but the friction pads it used turned out to be unsuitable regardless. I sourced a new clutch with a different tooth design from a different catalog and redesigned the clutch assembly around it to meet our requirements.

controller assembly: envelope & board architecture

I owned the full mechanical design of the controller assembly for the HVAA pod, from initial envelope allocation through ICD handoff to the PCB layout team. The assembly houses two functionally distinct boards (power and controls), a chassis-mounted fuse box, and an all-board-mounted MIL-DTL-38999 connector interface, all packaged inside a chassis constrained by shelf space on the pod.

The first task was carving out a usable volume on the pod. I was allotted space for two boards on a common plane, which set the basic form factor: the chassis needed to be wider than it was tall, and it would sit upright rather than flat to minimize its footprint on the shelf. That orientation decision cascaded into everything downstream: board layout, connector access, and how the fuse box would nest into the available depth.

The assembly split into a power board (power distribution/conversion) and a controls board (signal processing and decision-making logic) on the same plane. Because the two boards operate in electrically noisy (power) and electrically sensitive (signal) domains respectively, they needed EMI isolation from each other despite sharing the same plane and chassis volume, while still being tied together electrically via a board-to-board interconnect cable large enough that its routing had to be accounted for early in the layout.

controller assembly: EMI shielding

EMI protection had two layers: keeping external EMI out of the enclosure, and isolating the two boards from each other internally. For external shielding, I incorporated a standard EMI gasket groove into the chassis, sized to the spec of the EMI rope gasket I selected. Since the enclosure was split into two board compartments, I gave the internal dividing wall its own EMI groove as well, so the power and controls sides would each seal against a gasket independently.

The two boards still needed to be electrically connected to each other, which meant a hole through the center wall without breaking the EMI seal it was there to provide. Machining that hole with a 3-axis end mill wasn’t trivial: I set up the cuts so the center wall could be undercut with a keyseat cutter, allowing a hole to be machined into a face parallel to the cutter’s direction of travel, after digging into minimum shank diameter vs. cutter diameter tradeoffs to find a cutter combination that actually worked.

With the hole in place, I designed a feedthrough plate with an EMI gasket on each side, fitted with makeshift backshell attachments. Those attachments picked up an EMI shield braid, which in turn connected to the harness mounting to both boards, maintaining shielding continuity across the passthrough rather than leaving it as an open gap in the wall.

controller assembly: connectors & human engineering

Two requirements drove the chassis depth: a chassis-mounted fuse box needing dedicated volume (an initial rough depth target of ~3 inches before deeper design work began), and the EE team’s strong preference for all connectors to be board-mounted rather than cabled, to minimize interconnect cabling. The latter meant extending the depth on the aft side of the boards so each connector could mount to the board and the chassis simultaneously at a defined depth. MIL-DTL-38999 connectors being extremely well-documented made this part straightforward — the spec gave me everything I needed to nail the mounting geometry and clearances.

Connector placement wasn’t just a matter of fitting geometry. I applied MIL-STD-1472 (human engineering design criteria) to define spacing between connectors, ensuring the layout was serviceable and usable by a technician, not just physically feasible.

The design proceeded in sequence: define connector locations (geometry + MIL-DTL-38999 constraints + MIL-STD-1472 spacing), define fastener spacing in an array around the PCB outlines, package that into an ICD and hand it to the layout team to begin PCB design, and concurrently develop the EMI passthrough design for the interconnect cable. Once the ICD was in the layout team’s hands, the design went through several iteration cycles to accommodate needs surfaced by the EE team.

controller assembly: verification & integration

With the design locked down, I ran the full assembly through structural FEA against the applicable loads and did a DFM pass. Before committing to hardware, I also 3D printed the assembly, including a mock PCBA, to catch any fit issues that analysis alone might miss. That check came back clean, and the design was released to build and test.

One issue surfaced during integration: the D-slot for one of the D38999 connectors had been cut slightly oversized. It slipped through for two reasons: the connector itself visually obscured the slot in the assembly files, and interference detection in the CAD tools isn’t built to catch an oversize clearance cutout like that; there’s no actual interference to flag.

The fix: I designed an adapter plate with its own EMI gasket that mounted onto the existing chassis over the affected cutout. Since the original connectors were both board- and chassis-mounted, the fix had to preserve the same flush mounting surface as the original design, so I modified the chassis to accept the adapter plate, mounting it with screws and nuts. With that in place, integration proceeded successfully.

status

[Fill in current status, e.g. qualified and in production, or what’s next]