WebTV Galaxy

Galaxy brought the heat, storage and connectors of a small computer into a television gateway. WebTV’s HomeNet plan called for three satellite tuners, two hard drives for DVR storage and HDTV output. Inside the enclosure, the mechanical problem was to make the drive mounts, CPU shield, partitions and rear electronics work as one package.

I provided mechanical design and Pro/ENGINEER documentation through Mechanistic, working under WebTV’s direction. The work moved through layout studies, supplier feedback, prototype files and detailed DEV revisions. Roy Askeland coordinated the assignment; Frank Salinas supplied hardware-system requirements and review, Avijit Goswami performed the thermal analysis, and Mike Hiett managed the prototype effort with EMS.

A gateway for a planned home network

The September 2000 HomeNet material placed Galaxy at the center of a proposed network of entertainment devices. Early appearance studies explored an UltimateTV-branded enclosure with a broad front window and a central control. That exterior study and the later development chassis represent different parts of the record; the assembly views show the internal engineering more directly.

The mechanical discussions were already taking shape in March 2001. Roy asked me to begin component-concept discussions on March 5. On March 13, I proposed reviewing Galaxy case-closure concepts, and Roy said he would bring more specific thermal constraints. The messages show how the assignment was being framed before the later simulation and detailed review records. They do not establish that the proposed meetings occurred.

By spring 2001, the package had to accommodate two drives, a main board, a separate CPU board and shield, tuner boards, a smart-card reader and a power supply. Rear-panel connectors tied the electronics to chassis openings. Board heights, cable exits and mounting details were still being resolved while the mechanical model developed.

The parts list was still a planning tool. In May, the hardware team distinguished newly assigned part numbers from the updated preliminary BOM and noted that hardware quantities were still being gathered before entry into Agile. The later reviewed sheets likewise retained tentative and to be determined entries. Those records explain interface definition; they are not a final as-built configuration.

The open-chassis view makes the main relationship visible: storage and the CPU occupy the front of the package, with partitions separating them from the rear electronics and power supply. A change to one mounting detail could affect both assembly access and the route available to air.

Heat set the route through the enclosure

Avijit Goswami’s April 23, 2001 simulation evaluated a 150 W total load at 45°C ambient in an enclosure that had grown to 300 mm deep. Air entered across the hard-drive area and exhausted through the tuner, smart-card and power-supply regions. The CPU model included an extruded-fin heatsink, an intermediate aluminum spreader and an assumed grease or phase-change interface.

The model reported 28.3 CFM through the box, a 52.8°C drive case and 55.0°C air near the smart-card reader. Its CPU result was 84.0°C, labeled “junction/case” in the report. These are analytical results with specific component and airflow assumptions. The temperature contour and velocity plot preserve the original fields and scales.

The analyst singled out the smart-card location for further attention and recommended moving it closer to the inlet. That made component placement part of the cooling problem. The proposed location change remained a review action in the available record; the model alone did not establish the performance of a finished prototype.

Drive support and air partitioning needed different parts

The May mechanical review considered making the hard-drive bracket serve as an air partition. The notes rejected that combined function over a 1 g concern and retained a separate partition to direct airflow. Mounting the drives and steering air therefore became distinct design jobs.

The same review addressed bracket attachment, formed stiffening, grommets and ferrules. It also compared a vertical, rear-access smart-card arrangement with a horizontal reader beneath the drives. The vertical option was recorded as the lower-cost, lower-complexity direction, with mechanical follow-up assigned to me.

These choices connected the thermal layout to parts that could be mounted and serviced. They also left details to work through: the partition’s clips and holes, its fit around adjacent hardware, and the routes for power and drive cables.

The CPU package had to be assembled in place

The CPU enclosure combined directed airflow with EMI shielding. Its perforated top admitted air beneath the fan, while the lower shield and connection to the main board made the underside important too. The top assembly and underside view show both sides of that problem.

The July 27 DEV2.1 review exposed a practical constraint: CPU fasteners lay close to a fixed partition, so the subsystem could not be dropped into the chassis as a preassembled unit. The recorded direction was to build it in parts. The review also called for flanges around the enclosure and a better seal, leaving shielding details to resolve alongside assembly order.

Vent geometry continued to change. In August I sent a revised CPU top for review and then requested prototype pieces. A September hotlist still carried vent, fan-gasket and mounting changes. The records use different dimension conventions for the hexagonal openings, so they do not establish one final vent specification or a completed emissions result.

Tooling feedback changed the geometry under discussion

WebTV asked NT Vision to assess the CPU shield early, while its internal requirements were still developing. The May exchange proposed embosses and ribs for stiffness, folded sides in place of the initial box form, and multiple blanking stages for the dense hexagonal hole pattern. Pauline Soh’s response specified at least three blanking stations for the 174-hole pattern then under review. This was a tooling proposal, with final gauges and geometry still open.

The chassis base raised a related forming question. In June, Chuck Pokonosky at D&D Manufacturing warned that the quantity and placement of drawn standoffs could warp the base. He proposed alternative standoff forms or strengthening ribs. Frank welcomed the rib discussion while flagging the exposure of internal areas that the alternative forms could create.

The prototype-deviation drawing, transmitted in July, makes the consequences concrete. Selected domes were to be replaced with 6 mm standoffs; other dome locations were allowed 7 mm height with corresponding changes to panel openings. A local manufacturing accommodation could move the board and force an adjustment at the enclosure boundary. These proposals document the work toward a buildable package; they do not establish a production stamping yield.

From prototype files to the last recorded review

My July 7 status reported that prototype files had been released on June 28, with Mike Hiett managing the EMS effort. Flexible-partition prototypes had arrived; PCB prototypes were still expected. The July DEV list then recorded concrete assembly and fit problems, including the CPU fasteners, power-supply installation, connector alignment and partition details.

Revisions continued into August. On September 17 I described Galaxy as nearly wrapped up and asked to review the current CPU assembly once more. That is the clearest late status in the reviewed correspondence. It extends the story beyond the initial model and release package while leaving the final review outcome open.

Galaxy’s contribution was a mechanical development package carried through a coupled set of cooling, assembly and fabrication constraints. The record supports that work and its prototype milestones. Product shipment, full qualification and the eventual program decision remain unestablished here; later Xbox work has its own engineering history.

Source trail

• The April 2001 mechanical-design proposal and September 2000 HomeNet planning material establish the assignment and intended gateway.
• The April 23 thermal report supplies the model assumptions, results and analyst recommendation.
• May mechanical notes, the July 27 DEV review and the September hotlist record alternatives, assembly constraints and continuing revisions.
• Original architecture, CPU and base-deviation figures show the parts under discussion.
• Dated WebTV correspondence records supplier feedback, the June prototype-file release, August CPU revisions and September status.