A brand new Flux 3D printer arrived from Taiwan! This product is still very much beta, so I’ll leave it to others to nit-pick the many rough edges. My goal here is to give a quick look at some of the interesting technical details.
The Flux was a huge hit on Kickstarter, raising $1.6 million on a $100,000 goal. They missed their July 2015 estimated delivery date, but these days I consider any tech Kickstarter that actually ships anything, ever to be a success (so many don’t).
- Delta mechanism. Pretty standard Rostock style.
- Slotted belt drive. Cheaper than threaded rods, not necessarily worse for a small printer like this.
- Bowden tube feeder with plunger in the top. Hmmm, I’m not so happy about this because it adds slack when doing precise feeds and retractions. It also makes it hard to print tacky filaments like NinjaFlex. I’d much rather see a plunger right above the hot end with a short and direct filament path into the hot end, but that would add weight to the end effector.
- Filament spool storage in the top of the unit, Neat and tidy, but I personally don’t mind having an external spool since it lets you use any size and keep lots of different materials and colors ready to swap into the machine.
- Thick, metal, unheated build plate. It is very sturdy and so likely to stay planar. It rests lightly on rubber stand-offs with only gravity to retain it.
- Integrated camera that could potential be used to remotely monitor print progress.
- Pop up lasers and a turntable for 3D scanning. Looks like a standard system like the MakerBot Digitizer.
- Sports a land-locked Raspberry Pi and an ARM Cortex M3. Plenty of processing power, although a round-about way to get there.
- Auto-leveling uses load sensors under the build plate to detect the nozzle touching down. Seems to work well and solves some problems with other methods.
- Magnetic ball-and-socket joints on the linkages. This is an interesting solution to making universal joints, and they let you very quickly swap out the end effector.
The trolleys are driven with toothed belts driven by steppers hidden in the base.
There is a home limit switch at the top of each rail…
Nice linear bearings run the rails.
There is an electrical connection between the two sockets in the trolley that looks like it was added intentionally. There is no electrical connection between the sockets and the rails. The arms connect into these sockets and the sockets on the end effector with magnetic ball linkages.
The sockets are strong magnets, the balls are not magnets.
The spool rests on its side in the top of the machine.
The filament feeds though a little hole in the cover. There is an optical sensor in the hole and a spring loaded lever for loading.
There is a spring loaded groved wheel on a bearing to push the filament against a ribbed drive gear, which is driven by a stepper motor.
It looks misaligned here because the plastic cover normally holds the lever in place. The assembly works a lot like whpthomas’s pinch roller design.
From here, the filament follows a bowden tube down to the end effector.
Inside the end effector, there is a control board, 3 cooling fans, and the hot end. There does not appear to be an electrical connection to the sockets, so I guess all the power is coming down the control cable. When I saw the electrical connections in the trolley, I thought maybe they were sending the power and data though the rails and the arms, but maybe they couldn’t get it to work?
There is a control board in the top that is connected to the filament drive stepper, the filament sensor, and the homing switches. It is connected to the base using wire conduits that run down between the rails.
There is a control board in the end effector that is connected to the cooling fans and the hot end. There is a pluggable cable that connects this control board to the one in the base. I think it is the first USB-C cable I’ve ever seen in the wild.
There is a custom control board and a Raspberry Pi in the base.
The Raspberry Pi is completely land-locked. Its header pins 1-26 are connected to the custom board, as it the USB port.
I put on a serial tap and booted it up, but didn’t get anything after the “Decompressing Linux” message, so maybe they’ve disabled the console to use the tty for something else?
There is a beefy 6-sided steel build plate that is about 2mm thick.
The plate rests on 3 load sensors under 3 of the corners. These are used for auto-leveling.
To auto-level, the nozzle touches off each corner above the load sensor at least twice. Then it touches off the center of the build plate.
Out of the Box
Everything came very well packed. Build quality is high. Assembly consisted of only…
- clicking the magnetic linkages in
- connecting the bowden cable
- plugging in the hot-end control cable
- placing the bed plate into the bottom
You don’t need any tools, but the threaded end of the bowden cable is in a well, so a wrench or pliers makes screwing it in easier.
Overall assembly took maybe 2 minutes.
The box also contained…
- A TP-LINK TL-WN723N USB Wifi adapter so people who don’t have Wifi can connect to the Flux (who doesn’t have Wifi?)
- Glue stick. You put a couple of layers of glue on the bed to help with adhesion.
- A tube of lubricant.
- A scraper for removing prints from the build plate.
- A very minimal manual.
- Two spools of 500g each of 1.75mm PLA filament.
Once you download the controller software (Windows and OSX), you connect to the Flux over USB and either log it into your Wifi network or tell it to make a local Wifi AP if no network is available. Once that is done, you unplug the USB and all communication is over the Wifi connection. (I had to restart the software to get it to connect over Wifi.)
I had very bad luck getting the Wifi to work. I tried different APs, making the Flux be its own AP, and starting the Flux and/or the Flux Studio software in every possible order. I only was able to get the software to find the device after the Wifi setup step maybe 10% of the time with no pattern I could find.
FLUX Studio is pretty basic, but fairly complete. It feels a lot like MakerWare or ReplicatorG.
It seems to only be able to import STL files, so there doesn’t seem to be a way to print pre-sliced tool paths. This could be a problem for people (like me) who like to tweak their printing, but maybe not a big deal for average users.
It appears to use Slic3r for slicing and also has an unselectable “experimental” slicer option without any explanation as what it might do differently.