KPLAYER-6000

Kaleidescape’s KPLAYER-6000, codenamed Apollo, brought 1080p output to a player that could import DVDs and CDs and play from either a disc or the server library. The product kept the familiar white front, blue glow and one-rack-unit format. Its mechanical challenge was to make that inherited package work consistently as the electronics and production supply changed.

My contribution was mechanical integration and production support: coordinating revisions with Argyle Design, reviewing product specifications, supporting the door-spring change and preparing installation information. Argyle retained the underlying Apollo CAD during the launch work, and the work joined its mechanical design team with Kaleidescape’s hardware and manufacturing groups.

A new player inside a familiar package

The August 2007 requirements defined Apollo as the successor to Movie Player 2. The aim was improved deinterlacing and scaling to 1080p, while retaining disc import and playback, broad output compatibility and the established industrial design. Mercury, the smaller player being developed alongside it, had a different brief for secondary viewing locations. Apollo’s front door, optical drive, internal power supply and rack installation all remained part of the mechanical problem.

That continuity mattered on the factory floor. A familiar enclosure could carry earlier interface problems into the new product. The door had to return reliably; the baffle had to clear cables and fasteners; and the drawings supplied to production had to describe the configuration actually being built. The product also depended on front and rear ventilation clearances.

When better location exposed the interference

In May 2008, manufacturing found that the front door could hang against the bezel screen. Jon Cohn at Argyle checked the design and found a discrepancy: the CAD used a 0.750-inch hinge pin, while the purchasing reference called for a 0.625-inch pin. The shorter pin allowed the door to move in its supports. One sample could clear the screen while another would rub.

Correcting that discrepancy did not immediately solve the door. On May 22, John Lio reported that Michael Clader had assembled the longer pins with parts molded from the modified screen tool. The door was centered and its unwanted movement was gone, but it now hung up consistently. Better location had put the door in its intended path and exposed the insufficient clearance in that path.

The team tested the available responses. Adding a turn to a spring made the test assembly close every time, but Lio described the force as probably too high for production. Clader and Cohn considered thicker wire with fewer turns, while also discussing spring deformation, heat treatment and friction over time. The experiment separated the ability to push past an interference from the choice of a spring suitable for production.

By May 28, Clader recorded my agreement to order 1,000 springs from Peridot using 0.017-inch plated music wire and 4.5 turns. That decision moved the team from a useful bench experiment to a specified spring purchase, with prototype testing planned in parallel.

Sticking doors appeared again in the August 21 sustaining review, with Clader following up. The September 11 minutes marked that issue resolved, without identifying the corrective change or its validation.

The air dam was also an assembly interface

The internal air dam exposed a different kind of integration problem. On May 19, Clader reported that the captive-washer screw joining the right drive bracket to the baffle stood about 1.5 mm too high. He also found a pinched power cable beneath the baffle and suggested reversing the bracket tabs to make the drive removable without first removing the air dam. Those were three distinct concerns: fastener clearance, cable relief and service sequence.

I asked Argyle for revised files and offered to coordinate the changes. Gabe Cohn replied that the cable-relief change had already been issued in February, and questioned whether production was using the earlier drawing. He also challenged the washer assumption: the nominal clearance he calculated used a pan-head screw without that washer. The service-sequence suggestion remained a tradeoff because he expected the alternative assembly order to make the tight drive installation harder.

The assembled hardware, fastener specification and released drawing needed to agree. Richard Lane’s follow-up questioned the conflicting revision labels and who had actually supplied the metalwork package to the manufacturer. The geometry review had become a configuration-control investigation: a cable-relief revision could only help if it reached the parts being built.

Checking the specifications and readiness

The same discipline applied to the product sheet. When Linus Wong noticed that Apollo’s power, heat and weight matched Movie Player 2, he asked whether they had been measured or copied. Rob Wudrick confirmed that the power and weight had been measured; equal power also supported the same heat-output specification. I reviewed and supplied edits to the specifications as the datasheet moved toward release.

The published product sheet lists one rack unit, 45 W maximum consumption, 150 BTU/hr maximum heat output and 5 CFM minimum airflow, with a 5-30°C operating range. These ratings describe the operating envelope installers had to accommodate.

At the June 13 readiness review, Apollo’s required development and certification work was reported complete, with approximately 150 finished units and production running at 50 per day. This records launch readiness at a specific point in production.

The production record continued after that milestone. The June 26 NPI meeting moved Apollo into sustaining while Michael Clader followed up on manufacturing-test coverage and a board build whose connector population differed from the BOM. By July 3, the test fixture had been updated for the mechanical and connector changes, although the team was still checking its coverage. Launch readiness, repeatable manufacture and continuing support each had their own work to finish.

Changes continued after the first configuration

The SATA version was already being prepared alongside the initial Apollo configuration. The May 19 BOM release authorized prototype purchasing and still lacked a finalized SATA cable. Wudrick argued for separate assembly identities and a controlled ECO cutover so the two versions could be built and traced independently. In June, Vincent Di Lello sent Argyle mechanical-check files for the new Ethernet PHY, SATA controller and two SATA connectors; the fan connectors had moved to make room. A change of drive interface carried consequences for board layout, cable routing and product traceability.

Di Lello uploaded revision 1 of the SATA board files for an ECO on June 19 and revision 2 on June 24. The change then had to pass through a prototype build and bring-up. In August, the NPI record reported five units delivered, with drive qualification and chip-level issues still being investigated. Drive selection remained open: June notes say a candidate had already gone end of life, while August notes report failures on difficult discs during qualification.

By September 11, the minutes reported initial hardware bring-up complete for the TL9 SATA configuration, including DVD qualification and qualification of the 80 W supply. EMC work and separate TL10 builds were still ahead.

The later records distinguish different chip revisions and test states. An October 22 meeting reported that the TL9 SATA configuration had passed an EMI prescan while TL10 bring-up was pending. On October 29, the team reported bringing up ten TL10 PATA/SATA units: five had worked initially, while four had cracked resistors and one had solder shorts. Safety, EMI and thermal work for that configuration remained pending. The interface change continued through board, software, manufacturing and qualification work.

The December 11 sustaining minutes subsequently reported Apollo SATA EMI certification complete. That entry does not specify the chip revision or establish completion of the separate safety and thermal work.

Support continued after units entered service. In October, Adrian Fung’s hardware meeting record reported two Apollo field failures involving Q34 and described the electrical team’s change to R291, from 0 to 6.2 ohms, to move dissipation away from the transistor.

By November 20, Proworks had reworked the Apollo boards it held for Q34 failures, including its failed-board inventory; the minutes reported that this fixed some of the boards.

My work also extended to the information needed to install and integrate the product. I supplied updated top-view drawings for review in August, and the web ticket records their addition to the extranet. In January 2009 I requested the complete Apollo CAD package from Argyle to support an aviation installer. These are practical release deliverables: the physical product and the data that other people need to work with it.

What the platform made reusable

A March 2009 discussion about the next player generation illustrates the continuing value of the inherited chassis. Wudrick argued for retaining the existing hard tooling and changing the rear panel. His comparison cited a $27.92 hard-tooled KPLAYER-6000 chassis against an $84.62 soft-tooled KPLAYER-5000 chassis. The proposal treated Apollo’s established tooling and internal layout as assets to weigh against the flexibility of a new enclosure.

Apollo’s account is one of making a mature package dependable enough to build and support while its interfaces changed. A hinge-pin correction exposed a clearance problem. A baffle complaint exposed uncertainty about which drawing had reached production. A board respin required configuration control as well as connector clearance. My contribution sat at those boundaries, working with the people who owned the CAD, electronics, manufacturing and release information.