Glyph

Joining Avegant

Richard Lane, my former boss at Kaleidescape, recommended me to Joerg Tewes, Avegant's CEO. I joined on July 20, 2015, as a Senior Mechanical Engineer to help get Glyph into production and get its overdue Kickstarter deliveries moving. Glyph was the main reason I was hired and became the bulk of my work there.

Glyph combined headphones and a personal display in a 434 g (15.3 oz) headset. Lower the band across your eyes and it became a display; raise it and it returned to headphone use. I took responsibility for the whole mechanical bill of materials: getting the plastic and metal parts, springs, cables, pads and their interfaces through tooling validation and production ramp. Color, material, finish, print and graphics were part of that responsibility too.

An overdue Kickstarter commitment

Before I joined, I looked into the overdue Kickstarter deliveries and the growing backer frustration. The product design was established, but there was still a lot of work between that design and reliable production. I was coming in to help close that gap.

The campaign had raised $1,509,506 from 3,331 backers. The original hardware reward was a beta headset, estimated for December 2014; Avegant subsequently committed to delivering a consumer product instead. When I arrived, that original delivery estimate was already seven months past.

I worked on as many fronts as I could at once. Assembly and fit, cable binding, dust contamination, liner behavior and comfort could all hold up shipment. Pilots and production builds exposed some problems; testing, people wearing the product and customer feedback exposed others. A change to one part had to work within the rest of the assembly, be practical for the supplier to make, and go through the validation it needed. That work continued as production ramped and deliveries began.

It was pandemonium. Testing, supplier questions, tooling changes and production problems were all moving at once. I had a lot of plates spinning, moving between the immediate build problems and the changes needed for the next round.

Fitting two optical engines into the headband

For my mechanical work, the challenge extended from the optical assemblies to the person wearing them. Within each engine, the DMD, illumination optics and eyepiece had to maintain their geometry. The two engines also had to stay aligned with each other as the wearer adjusted their spacing, or interpupillary distance (IPD). The chassis supported those relationships inside a headset that flexed, telescoped and moved between headphone and display positions.

I supported the triplet-lens bonding process early in the program. Chris Westra was taking point on the optical work. I came in to fix problems in both bonding and optical inspection and alignment.

Getting the mechanical BOM ready to build

My tooling work ran from design through ramp. I worked with Chris Westra, Misha Young, our contract manufacturer Intretech and the component suppliers to turn tooling samples into parts we could build with. That involved inspecting dimensions and appearance, checking how the parts fitted together, getting tool changes made and deciding what could proceed into a build.

Keeping a build moving meant knowing which parts were ready and what still stood between us and a complete set. In late August 2015, we were working toward material readiness for 10 sets before August 26 and 135 sets before September 3. The headband BOM tied those quantities to responsible people, first article inspections, review status, whether a part was usable for engineering validation, and the actions needed from the supplier. A part could be usable for that build while still needing a tooling correction for production.

The outer-headband tool

Of the roughly 40 tools I was responsible for on Glyph, the outer-headband tool was the trickiest. The part, 425-0003, was a sweeping arch nearly 200 mm across, molded in SABIC EXL1414 impact-modified polycarbonate.

Moving to a hot-runner setup was a big part of getting the outer-headband molding to work. I ended up tuning the process parameters to produce the right as-molded part weight. That gave us parts that could eject and then cool into the right shape and size, repeatably. This kind of process tuning is familiar work in plastic molding. What made the headband difficult was how sensitive the whole part and tooling system was.

Keeping the wider build moving

I delegated as much of the drawing and documentation work as I could. Leonard Pang and others helped with drawings and revisions. That let me protect time for troubleshooting, supplier work and the decisions needed across the mechanical BOM.

That work extended across the BOM: spring indexing, gimbal holes and pins, sliding and bearing surfaces, snaps, screw features, texture and finish. We had to distinguish a drawing or measurement problem from a tooling or processing problem, agree on the correction and check the next parts. Chris, Misha, the supplier engineers and I worked through those reviews together.

The first question with a fix was how much validation it would take. Then came what the supply chain and fabrication could support. I coordinated on the fly, trying to make the best use of time and money. I worked for Chris Westra and Allan Evans, who approved major expenditures. Those practical limits shaped how much we could change, and when.

I've always been an avid photographer, and I use photography to communicate technical work. At Avegant, photographing parts, assemblies and the work around them was part of pursuing good parts at volume and acceptable yield.

Designing the assembly and inspection fixtures

I designed all of the assembly fixtures, along with inspection fixtures for parts such as the outer headband and inner liners. The geometry we were asking suppliers to hold had to be reflected in how we supported and located the parts for inspection.

The outer-headband fixture followed the long arch and located it at labeled positions. The liner fixture supported the smaller, ribbed part in a shaped nest.

I wrote individual inspection procedures and acceptance/rejection criteria for critical parts, including the headband and liners, arm inners, ear covers, bezel rings, shutters, button plates and backers, ear pads, bushings and faceplate. I also created the master gap specification (795-0002), defining target and allowable gaps at the assembled interfaces. The fixtures, procedures and gap plan connected the part drawings to how we would inspect parts and judge the assembled product.

Defining and running qualification

I designed and managed the reliability testing program across Glyph. I worked with Intretech on how we would qualify the product and validate changes. The work covered the assembled headset's moving interfaces, environmental exposure, drops, buttons and connectors, as well as the durability of coatings, printing and packaging. It involved test samples and fixtures, operating conditions, inspections during testing and functional checks afterward.

The testing program at a glance

The spring study was one part of this larger effort: its April compilation records 950 measurements across 19 assemblies. It covered three spring groups, ten head widths from 100 to 190 mm, and telescope extensions from 0 to 40 mm. The interactive chart later in the story opens one of those five settings: 190 measurements at 20 mm extension.

We planned the September reliability allocation around a pool of 20 units, assigning samples across the different tests. As the design and process changed, specific fixes needed their own validation. In April 2016, after initial shipping had begun, I had cable-abrasion fixes in validation on 20 units for headband lifecycle testing and another 10 for drop testing. Short-term chassis machining, cable rework and a revised routing scheme were being evaluated alongside longer-term tooling and harness changes.

By April, the six-variant cable comparison was complete and we had rejected the splint approach. Testing was helping us decide what to change in the product already being built.

Qualification continued through the ramp. I had to coordinate the parts and sample configurations, the supplier changes and the testing needed to put a change into production, while backer deliveries were still underway. Comfort brought us back to people wearing the assemblies we were making.

Making Glyph wearable

Taking the load off the nose

I couldn't wear Glyph for more than a few minutes without pain. We kept hearing that people's ears hurt, and the silicone nosepads left deep, angry marks on many test subjects that took a long time to disappear.

Radius Product Development supported the headstrap work.

The headstrap was a late addition. I proposed it, got it approved, tested it and brought it into production relatively quickly. Radius developed support concepts and mockups during that work. I added receiver features to the earcans for quick-release buckles, so the strap could be removed and Glyph could still be used without it.

With the strap attached to both earcans and adjusted over the crown, it took almost all the load off the nosepiece. The nosepiece then mainly located the unit, while the top of the head carried the weight. The strap also added another contact that helped keep the display in place.

It was a hold-your-nose solution. The strap was big and bulky, caught in people's hair and was tricky to adjust correctly the first time. It basically turned the headphones into a hat. We accepted it reluctantly, late in a long and frustrating engineering effort under pressure to deliver the overdue Kickstarter product. Once people got the hang of it, though, it became essential for longer use. It was the comfort fix I felt had the most impact.

Keeping it optional left another constraint in place: the earcups still had to grip hard enough to hold Glyph without the strap. I reduced the clamp as far as I could while preserving that use. Even then, it was too much for me to wear comfortably for long, with or without the strap. Taking weight off the nose made a substantial difference; the ear pressure remained.

Tuning the pivot feel

I also specified a damping grease for the headstrap pivot to get the feel and resistance right. I wanted something like the controlled movement of an expensive volume knob. Pivot torque testing was part of that work.

Measuring the assembled clamp force

Clamp force changed with both head width and telescope position. We measured the assembled headset across those settings while comparing spring and liner configurations. Jill helped with assembly and testing during the spring-and-liner work. At 150 mm width and 20 mm extension, the recorded 1.6, 1.4 and 1.2 mm spring groups averaged 11.7, 11.0 and 9.6 N. The curves show why a single force number could not describe the fit.

Developing the earpads

The earpads were a development effort in their own right, balancing contact pressure, comfort, traction and appearance.

I tore down numerous headphones, obtained vendor samples and bought replacement earpads for a wide range of models. These gave me a broad set of constructions to examine while designing and prototyping Glyph's pads. My favorite Bose earpads were a reference for the foam type and combination.

I built up a group of Glyphs with the different earpad variants and had colleagues try them and provide comments. Nearly everyone liked the dots and the low-pressure foam, and I used that feedback to make the selection. We chose the normal dotted surface, shown in the finished earpad photograph, rather than the gel dots or silicone netting.

The inner liner and its fixed constraints

The inner liners were among the most difficult parts I worked on. These two-shot molded parts had to move within largely fixed interfaces: the surrounding parts, travel and extension distance left little room to change the design.

During flex or twist, the liner and its mating flange did not deform together. A small gap would open at the edge, like a coin purse: the dreaded pucker. I tried incremental changes, but we had to live with some puckering and deformation. The product shipped with those compromises still in place.

Trimming presented a separate production problem. The fixture technique was difficult to reproduce, so we ended up trimming the liners by hand. One operator had mastered the job, and she was the only person who did it.

Movement, reliability and dust

Cable routing through a moving assembly

The cable could work initially, then develop a kink after repeated cycling and bind soon afterward. Adding stiffness could help it resist buckling, but the bundle still had to fit and move through a confined path. Protective tape could curl at its edges and increase the effort needed to telescope the arm. Local stiffness, bundle size, friction and routing had to work together.

I spent a great deal of time working through those combinations: splints, different jackets and wires, changes to the looming specifications. Careful design of experiments, lots of changes and lots of cycles.

With the splint approach rejected, we were working on three fronts:

Mechanical ECOs for the surrounding parts were released while the alternative cable samples remained a separate development effort. The practical task was to keep improving the assembly in production while evaluating what should replace it.

The work carried into production through careful harness assembly, inspection and system assembly. During the short production run, I instituted a mid-build "break-in" cycling station to test and validate the process and support ongoing improvement. I specified cycling 20% of units; in practice, we only did one in ten. No kink meant pass and continue through assembly.

Electrical checks and physical condition

Electrical continuity did not tell the whole story. Of 17 cables returned from the production line for supplier analysis, 11 passed the electrical retest, five had open or short circuits, and one connector was too damaged to test. Some of the electrically passing cables still showed visible jacket or connector damage. Inspecting the physical condition answered a different question from checking whether the wires still conducted.

The work extended into retention, strain relief, adhesive placement and assembly instructions. We were working on the harness and the parts around it at the same time: modifying routing and wrap, protecting contact points and changing the chassis where it could damage the cable. Those changes needed to work together in the assembled product.

Pinching at the optical chassis

Some cable damage came from mechanical pinching and impingement during normal use, or was present straight out of assembly. I spent a great deal of time specifying secondary operations and trimming of the optics chassis, and documenting acceptable surface conditions and zones after injection molding.

Dust after assembly

With a disassembled Glyph, you could feel the air move as you extended and retracted the arms. Extension left more internal volume; retraction left less. Cycle it quickly and it pumped air through the gaps, carrying ambient dust toward the optics.

That meant a clean optical subassembly could become contaminated inside a nearly finished product. The optics-module assembly had its own dust problems. Cleanroom conditions and procedures largely brought that source under control, and every optical subassembly was inspected, with affected assemblies sent for rework or scrap. Some particles also came from materials inside the product. Around the folding mirrors, blackout-tape trimmings were a source of particles. The team's change to painted mirror edges made cleaning and rework easier.

I tightened the fits and tolerances between the stationary headband parts and the sliding earcan parts to reduce the gaps. I also added Kapton tape around the optical module to reduce the openings for particles. We never completely eliminated the pumping.

I changed the test order as well. Final optical checks moved into the final clean-area functional tests, after all the other physical functional tests. The optical inspection then checked the condition left by the operations that could contaminate the unit.

Diagnosing returned units

Diagnosing and mitigating field failures took a substantial amount of my time. A dead headset, a dark display or missing audio was a starting symptom. The investigation could take us through the mechanical assembly, cables, electronics and optics before we understood what needed to change.

Reproducing the reported problem

One returned unit was reported to have no left-channel audio and static. The March 2016 bench report found audio working in both analog and HDMI modes, but reproduced the static. Its recommendation was to check the twist and wrapping of the cable bundles entering the input board. The reproduced problem was narrower than the original complaint, and it led back to a physical detail of the assembly.

Production board analysis showed why similar display symptoms could require different fixes. In November 2015, the team restored a striped-display board by cleaning an oxidized DMD connector pad. Other boards worked after shorted components were replaced. In one repair, a 33-ohm resistor had been replaced with 180 kilohms; correcting it brought the board back. These were specific diagnoses and repairs, rather than one explanation for every nonworking display.

My mechanical work overlapped those electrical investigations. In April, cable wear at the chassis was suspected in two MCU shorts, and I was responsible for the mechanical response. That work fed into the clearance, protection and routing changes described above. Cycling, electrical checks and teardown each told us something different about the resulting assembly.

What the return records captured

A return-analysis log covering December 19, 2016 through January 12, 2017, during my final weeks at Avegant, gives a useful snapshot of the range of findings. It contains 292 observations across 290 recorded serial numbers: 220 observations marked failed and 72 marked passed. Two serials occur in both groups. The failed observations break down by their recorded primary code:

These are observations in one return-analysis log, not a failure rate for all shipped Glyphs. A unit could have more than one symptom, so its primary code was only part of the diagnostic picture. The mix shows why the work extended beyond whether a unit switched on: optical contamination, physical retention and electrical behavior all needed attention.

Factory corrective actions

A separate factory report records a revised mirror carrier entering production and a requirement to hold products in the QC room for seven days before first-article inspection and shipment. It also records supplier corrective-action follow-up for PCB via deviations and confirmation that a new shutter mold could meet its 90–93 degree bend-angle requirement.

Finish, graphics and personalization

Getting the appearance into production

I created Glyph's master color plan (795-0001) and graphic plan (795-1000). These defined the color, material, finish and graphics requirements across the mechanical BOM, linking the parts to material and coating specifications, color-chip references and inspection procedures. Glyph's look was already established; carrying it across different parts, materials and coatings took judgment.

The specifications I created were a necessary line of defense across the vendors in our distributed supply chain. CMF, graphics, gap and individual-part specifications worked together, and the part drawings referenced the cosmetic specification. They gave me a defined requirement to return to when discussing a part, an interface or a finish with its supplier.

I used measurements wherever possible, and always my own eye. Perfect color matching across all those adjacent surfaces wasn't realistic. The CMF and graphics plan helped me navigate the compromises and define acceptable bands for production cosmetics.

Customer artwork and employee units

My work also included customized aluminum discs for customer-supplied artwork.

For the employee commemorative units, each person chose artwork for the small aluminum center disc. I delivered the artwork to the vendor and coordinated getting the personalized parts made. I also designed the team-name artwork and arranged for it to be pad-printed on the right inner ear-cover bushing of those same units.

Shipping while the ramp continued

On January 28, 2016, Avegant announced that shipping was beginning. By February 3, the company had sent more than 200 tracking numbers, with yields improving and the line still below full rate. On March 3, it reported about 1,000 units shipped.

Through that spring, I kept working on the mechanical problems that came with the ramp. In April, cable-abrasion changes were still in validation, and I was working on clamp force, liner tooling and another cable supplier. The Amphenol work included defining the validation needed for its samples and coordinating that work with Intretech.

Delivery depended on work across the company too. CE certification held up international orders, and engraving added another step to fulfilling customized orders. On June 2, Avegant reported the Kickstarter units fully deployed.

Kickstarter fulfillment was a shipping milestone, not the end of the engineering work. The August clamp-force comparison still separated liner material, paint state and spring diameter. By December and January, the return-analysis work described above was feeding another set of findings back into the product. Testing, production support and field diagnosis continued through my final months at Avegant.

Leaving Avegant

I worked at Avegant from July 20, 2015 to January 27, 2017. When I left, Chris Westra shook my hand as I was walking out the door and said, "Thanks for fixing all of my problems!" Coming from Chris, that was high praise. It meant a lot to me.