Not too long ago I disassembled the V3 Starlink. Why? Because I wanted to see what was inside, and how I might be able to make a more “mobile” enclosure for the smaller, lighter “Dishy.” Here’s what I came up with.
It turns out that the active components of the Starlink V3 dish are actually quite compact and lightweight. As you can see, once “liberated” from the stand, this is all that’s left:
Not much to it! This is panel is only about 1cm thick, and weighs just shy of 3lbs.
My objectives were to make a new enclosure that is:
- Small, not much bigger than the panel itself
- Able to be mounted to a window/sunroof
The problem is, the “panel” doesn’t really have any good way to attach to an enclosure. In the SpaceX enclosure, the front plastic cover is ultrasonic welded to the back part, so they’re essentially one piece. The back of the PCB is supported by small plastic standoffs as you can see in this picture:
I considered a number of different options to make an enclosure, and ultimately decided on somewhat of a “sandwich” that the panel itself would be in the middle of.
I started by taking measurements of the perimeter of the panel, including the perimeter “channels” and PCB layers, and came up with a rough model of the unit itself:
As you can see, it’s not very detailed, but it’s enough to start working on an enclosure. On this model, I marked “safe zones” where there weren’t any SMT components (DBF, resistors, etc) and thus could be supported from the back with studs like the above.
I went through a number of different rough designs, but ultimately ended up with a two-part design that utilizes a backshell and front cover, with a tongue-and-groove assembly to make the housing waterproof. Here’s what it looks like:
With this design, the “panel” sits inside the back shell, supported by the perimeter shroud as well as the plastic standoffs in the “safe zones.” I added a channel for the Cat6 cable, which will be sealed with silicone and a grommet.
The front piece covers the full top of the V3 panel, adding about 2mm of plastic on top of the existing antenna face. I was initially concerned there might be some RF attenuation, but it doesn’t seem to be a problem.
To attach the front side to the back shell, the “groove” is inserted into the “tongue” on the back shell, and screws around the perimeter go through both the back shell and front shell to hold it together. You can see this design in the section analysis:
The teal part is the back shell, the blue part is the front shell, and the yellow part is the Dishy V3 panel. On the outside, where the the red circles are, I left a gap for a 2mm rubber seal, which will compress and weather-seal the front to the back when put together.
I was feeling pretty good about this design, but then realized I had no (good) way to bring it to life. This thing is big: 525mm x 473mm. I couldn’t make that on a 3D printer, at least not any of the ones I had available!
I looked around a bit and determined the best option would be to have it “printed” on a large-format SLA. But, this has its own problems: regular SLA resin is usually only rated to about 50-60C deformation temperature, and although it’s possible to get it made in Nylon, that costs big bucks (more than the dish itself!).
Because the dish is passively cooled, and because I designed the whole enclosure to be totally sealed, heat was going to be a problem. The dish itself is fine to something like 90C, so the low-temp resin wasn’t going to work.
Ultimately, I decided that I probably wouldn’t actually need it to be weatherproof. I’ll not be using it outside often anyway. So, I decided to add some venting to the hot section of the panel (near the CPU) and use the lower cost resin. Here’s what I came up with:
Looks kind of cool? In the final design, you can see 8 circular features around the perimeter of the enclosure as well. These are friction-fit holes to use for mounting later.
Feeling pretty good about this design, I sent it off to be made. A week later, it showed up. I was a little nervous about my measurements. Luckily, it fit like a glove.
It works! Only one screw-up: I forgot to add tolerance to the tongue-and-groove mating, which made the fit REALLY tight. I’ll have to make a new front piece at some point.
Because I didn’t care so much about waterproofing anymore, I also prototyped a “surround” that would replace the front cover, only holding the panel in by the perimeter. I chopped this up into pieces and was able to FDM print it:
Seems to work fine too.
With everything fitting well, I got to making some mounts. Elon said:
Much too big? I guess that depends what your tolerance for “big” is!
I whipped up some friction-fit suction cup mounts:
Printed them out:
And stuck it to the (sun)roof:
Turns out the upper sunroof has a metallic coating that does not pass RF. I had tested the V2 dishy in motion, but always in the back window – it was too big to fit up top. Once I moved it to the back window, it worked just as well as expected.
All in all, it was worth the effort – I ended up with a thick (15mm or so) placemat that can get high-speed, low-latency internet just about anywhere in the world. Pretty hard to believe.
So how does it work? Pretty much perfectly. As long as the dish has a relatively clear view of the sky, even laying flat on the ground, it seems to come online pretty quickly. Interesting, after the “motors stuck” alert becomes active about 5 minutes after startup, the connection seems to improve quite a bit. I’m still blown away by how well the phased array can track satellites – even picking up the panel and moving it around, it rarely loses connection – even at some pretty extreme angles.
It would be neat to see SpaceX release a version like this, where the panel itself is either detachable from, or without a base at all. This would be perfect for RVs, boats, trucks (and, um, other vehicles).
As a next step, I may make a base for the panel that incorporates the position motors – a kind of “cradle” that allows the panel to still be used in a stationary, motorized configuration when desired.