Bazooka (Base Station)

Bazooka was the wall-box base station behind Noon’s removable lighting controls. It terminated mains wiring, carried the power electronics and provided the mechanical and electrical interface to the Sativa and Elvis control modules. Waldo supplied the surrounding chassis and faceplate. Those pieces met at one installation, but they had different engineering owners.

I was brought into the program for mechanical recovery: establish focus, address the immediate failures and define useful short- and medium-term progress. My Bazooka responsibility was the Locoroll-side mechanical and reliability lead, spanning interface problems, requirements, test definitions and supplier handoff. Bel’s electrical and detailed product engineering, Function Engineering’s tolerance work, and FIH’s controller and qualification work were part of the wider effort.

A small enclosure with several jobs

The base had to fit a crowded wall box, retain a removable control module, make its electrical connections and disconnect the line when the module was removed. The metal front housing, rear plastic enclosure, PCB, connector interfaces and external TEST button therefore could not be developed independently. Wire exit position and stiffness affected installation; mechanical travel affected switching; the mating metal surfaces also participated in the grounding connection.

The requirements were moving. A June 2016 baseline called for 650 W; the May 2017 PRD called for 850 W. The May document also specified a mechanical air-gap disconnect with at least 3 mm separation and a 10,000-cycle life requirement. These were design requirements to satisfy, rather than measured results of the recovery work.

Giving the air-gap mechanism a controlled reference

The February issue list records two connected problems: the air gap did not consistently make electrical contact, and TEST buttons could stick instead of returning. A request to restore the square TEST shape was already present at that stage. The problem extended beyond the visible button to the guiding surfaces, contacts, spring and board geometry behind it.

The April 6 redesign proposal brought those functions together in a contactor guide fastened to the PCB. Plated contacts on the board would be bridged by a formed-metal shorting bar; the proposed bar’s compliance accommodated variation in the contact stack. The guide contained the return spring, guided movement and shorting components. This gave the mechanism a common mechanical reference close to its electrical contacts. It also required changes to the board and rear housing, including relocated wire exits.

By May, the EVT2-to-DVT action review described a new air-gap design, removal of the large PCB hole and lower rear wire exits. July prototype quotations and the later tooling and spring photographs show that the physical details continued to evolve. The April illustration remains a concept-stage view; the surviving release trail does not establish that every detail in it reached the final assembly.

A July 27 foam-spring study compared extra-soft slow-rebound PORON with soft and firm alternatives, using highlighted supplier stock sizes and stress-strain curves. This made the material response part of the spring investigation. A later square-coil concept image shows another arrangement under consideration. These records document alternatives; they do not identify one final released spring configuration.

Retention without an over-constrained fit

Bazooka’s drop-and-slide connection to the Waldo chassis had to feel secure while still allowing assembly and removal. Function Engineering’s four-hook analysis in April identified a basic constraint problem: making all four hook locations equally tight left no reliable choice of which features controlled X and Y position. Manufacturing variation could make different hooks compete to locate the same part.

The proposed remedy separated locating features from features that needed clearance. The hook-section comparison found another coupling: movement in X changed how tightly the chamfer wedged the assembly in Z. A flat bearing surface with a separate lead-in would reduce that coupling. Function Engineering also checked cover position against the remaining hook engagement, treating retention as an assembly stack rather than a nominal gap on one drawing. These were attributed analyses and proposed geometry changes, not a universal clearance adjustment.

Grounding and finish added another constraint. BAZ-41 required conductivity between Bazooka and the chassis, while the design review flagged an excessively large laser-cleared area and wear at the mating edge. The interface needed controlled contact and controlled removal; an appearance or fit decision could affect both.

Colin Davis’s retained native Y-direction calculation also examined the spacing between Bazooka’s locating teeth and the chassis slots. It calculated a possible 0.30 mm interference and proposed longer slots at three locations. The workbook is undated: it establishes another analysis state, without proving when it followed the tracker or whether the proposed slots reached the hardware.

Testing the complete interface

The EVT2 dust-test matrix recorded difficult removal for all seven Sativa and all seven Elvis samples at the Bazooka interface. The design review identified large square dimples as a likely source of surrounding deformation and proposed smaller circular dimples at the same depth. A later entry records a passing check with the control module, Bazooka and Waldo combined. The records establish those observations separately; they do not identify the dimple proposal as the sole cause of the later pass.

That distinction matches my recollection of the work: multiple dimples and assembly fits went through iterations, including reversals. Click, axial-pull and side-pull setups made different aspects of the interface observable. The useful question was whether the assembled system could engage, remain retained and be removed under its intended conditions, not simply whether one loose part met a nominal dimension.

I also worked on turning requirements into executable tests. Comparing the May PRD with my May 31 edit shows added methods, limits and durations for operating-temperature cycling, humidity, thermal shock, drop and vibration. The same operating-test conditions also appear in the earlier requirement baseline. Some fields remained open. Defining those conditions was a necessary engineering deliverable; it did not itself demonstrate a passing product.

The ten-year operating-life requirement became a point of dispute. I challenged it using test results, DFMEA and validation failures. The contemporary action review records the related concerns: it had not been communicated at kickoff, component vendors would not guarantee it, and the mechanical design had not been developed around that life. The recovered record supports the challenge, but not a formal deletion or a completed ten-year qualification. An EVT2 EMI pretest failure also remained a vendor action, with certified third-party testing requested.

The later bench record also contains pop and fuse-blow evaluations. Its indexes distinguish exposed-board baselines, in-wall tests, old and new TEST-button configurations, and different fuse ratings. The retained footage makes those test arrangements visible, but the sampled records do not establish a completed qualification or a single change that resolved the electrical and mechanical concerns.

I ran the reversed-polarity test series to investigate an installation fault that made the fuse fail violently, rather than simply trip. In that series, the destructive failure repeated every time. The electrical engineer believed the part was resettable, but the failures I observed destroyed it. I investigated the effects and possible mitigations without reaching a verified explanation or fix before the work stopped around the company’s closure. This is my recollection of the investigation; it does not establish a final fuse configuration or a released remedy.

A separate catastrophic failure involved two adjacent pins shorting under certain conditions and killing the unit.

Making the supplier handoff inspectable

The June handoff work connected part identity to usable engineering data. The CAD naming guide required model filenames to match Arena item numbers, CAD assemblies to follow the BOM structure, and drawing titles to match Arena item names. The mechanical assembly tree connected the housing, PCB, interfaces, button and bracket rather than leaving those relationships implicit in a collection of supplier files.

The associated redlines made inspection more explicit: metric dimensions, reference planes, numbered dimensions, designated incoming-inspection features and identified process-capability dimensions. Bel’s detail drawings and Function Engineering’s redlines retained their own authorship. My role included coordinating that data into a consistent Locoroll handoff. The June tracker still contains open entries, so it is a record of release work in progress rather than proof that every drawing was locked on one date.

The retained incoming-inspection records show why those details mattered. Their photographs mark surface and edge conditions around openings and retention features of the metal housing. They preserve concrete examples for review, while leaving the final disposition of the inspected material unresolved.

The later package checks show why receipt of files was only one step in that handoff. June questions still identified missing part mappings; July checks could read individual parts while retaining filename and assembly-structure problems. Those records document continuing release coordination, rather than a completed hardware release or shipment.

Following the part through the factory

By August 31, my proposed schedule for Best followed the metal front housing through casting, de-gating, de-flashing, drilling, tapping, CNC machining and finishing to final inspection. It exposed cycle times, working hours, staffing assumptions, carryover and the handoff between continuous work and batch operations. Colored cells identified limiting and non-limiting capacity. That made the schedule useful for discussing where parts would wait and which operation constrained the next batch.

The retained earlier sheets show different batch and staffing assumptions beside that staged plan. Comparing those alternatives made the handoffs and waiting time visible without confusing a scheduled shipping date with a completed shipment.

The August photographs show tooling work; September images show base-station samples, button sketches and return-spring arrangements. They place the mechanism work alongside the factory plan without turning proposed output into shipped units. The documented contribution is the recovery work itself: making interface failures observable, developing and checking changes, defining qualification conditions and coordinating the data and capacity needed to build the product. Final mass-production yield and shipment quantities remain outside the verified record.

Source trail

The account draws on the February Bazooka issue tracker; the April air-gap proposal; Function Engineering’s April fit and retention analyses; the May PRD and May 31 edit; EVT2 qualification records; June BOM, naming and drawing-review material; July prototype quotations; and the August 31 factory-capacity model. The photographs retain their dated sequence and design-stage captions. My recovery-role recollection is recorded separately from those contemporary documents.