Spit and Polish
Hello again. This time round we’ll be sharing a couple of the final touches to the Beast design process.
- A look at the refined design of the BeastTile front panel - the primary point of customisation for the system.
- Strength testing of the chassis to determine how high we can stack.
- A quick look at a refined design of the power hub.
By design, the front panel, or Faceplate as we’ve come to call it, is intended to be application specific. That is, it’s specially designed according to the particular devices, layout and cable-management required for a particular use-case. Until now, the Faceplate design has been something of a place-holder, while we developed the design of the underlying infrastructure (the chassis at the core of the structural system, and the power and networking hubs as the electrical systems). Now that we’ve settled on a release-worthy chassis and hub design, we got around to finalising a detailed Faceplate design for our particular application.
The first Beast revolved around the Raspberry Pi Model B, and so in keeping with tradition the Faceplate shown above is based on the same. This is the very first design of the Faceplate, which illustrates one particular configuration. Hopefully, more configuration will follow.
A couple of details to note about the design above:
- The cable openings are arranged within the footprint of the RPIs, so they are hidden from view.
- Once the connectors are through the Faceplate, the cable can then be led to a more suitable location via the slotted cutouts.
- By wedging the cables in these slots, they are held in place without cable ties or other additional components.
Just a small demonstration of how laser cutting can really be exploited to produce feature rich components that improve functionality without significantly increasing cost. The same applies to many other CNC--controlled manufacturing processes.
Having reached a near-final design it seemed prudent to prove some concepts by physical functional testing. From a mechanical perspective, the biggest question was how many BeastTiles can be stacked on top of one another before the bottom one fails.
So we built the test rig above, which enables us to apply a controlled load to the BeastTile, in the same mode as it would experience in practice. Essentially, it’s a vertical channel that keeps the tile vertical with a loose platform that slots into the channel and rests on the top edge of the tile (as shown in the image below).
Since we’re super-scientific and organised, we decided to use a 17.5 l paint drum as our mass-piece, which translates into a mass of approximately 20 kg. For now, approximately is good enough for us. Working on an expected average tile mass of around 2 kg, this test confirms that the tile can be stacked at least 10 tiers high. That’s well over our original requirement spec of 2 m. So that’s cool.
With the original Beast we experienced some SD card corruption that we attributed to poor quality output of the USB power supplies, particularly during power cycling. So we ran some preliminary stress-testing of the USB power hubs while under load with six RPIs, repeatedly cycling power to the hubs at various stages of the RPI boot process. Despite our best efforts to mess things up, we observed no booting issues, and certainly no SD card corruptions. Win.
Power Hub Design Revision
The first prototype of the USB power hub proved its performance nicely, but the wiring process was a bit too time-consuming and finicky for our liking. It required soldering individual segments of insulated wire (some of which are shown in the image below) between the DC jack, USB chargers and the power switch. For version 2, which we’ll prototype later this week, is a far simpler assembly, using two continuous copper wire rails - one positive, one negative - that span all four charger units, soldered to each. All internal wiring is thus consolidated into two rigid components that dramatically simplify the assembly process. Another benefit is that this approach doesn’t require dismantling of the car chargers at all. In terms of insulating the positive and negative rails, we’re going to play around with heat shrink and paint-on insulation gunk, expecting that heat shrink will win out.
Also note the informative engravings for the input power polarity and power switch on-position.
And that’s it for this one. Thanks for stopping by.
Any questions? Chat to us on gitter