A centrifugal turbine similar to those in real vacuum cleaners. The suction can be felt starting at c.a. 1800 rpm, becoming quite substantial at 5000 rpm (4 mbar). It was tested at 10700 rpm (the max I can get with my motor/power supply) and it didn't explode (take care when doing this!). Thick vanes provide even air channel inside the turbine. Overbent vane ends are supposed to improve suction and reduce noise by Coanda effect (doesn't work ;-). Thick vanes are hollow to reduce weight (thus allowing higher rpm), printing time and material consumption. The drawback is that water can get trapped in them and create imbalance, so the turbine is not suitable for water. (measured suction is 4 mbar at 5000 rpm. This is just slightly below theoretical value of 4.25 mbar as I have calculated. So, my coanda-aerodynamic-addition turned out to be complete bullshit! It is no more than a regular centrifugal turbine.) Outer diameter is 100 mm. Discs are 1 mm thick (with a slight thickening near the motor attachment), vanes are 10 mm high. Inlet hole diameter is 20 mm. There is a small hole in the center, which is to put the turbine on a standard 3.2 mm/1/8'' shaft (common for Dremel). Parametric FreeCAD v0.15 model is available, so all sizes can be changed at will.Before printing, adapt it to your motor. To do that, open the sketch "base" in FreeCAD model (TURBINE/Revolution/base in model tree) and edit it accordingly. When slicing, set the number of perimeters to something very large (e.g. 400). This is to force the base disks being printed in circles, in order to maximize its strength against being ripped apart by rotation. A concentric fill can be used instead, but in Slic3r 1.1.7 the bridge detection algorithm messes the bridging layer completely. When I printed it, the supporting center post was knocked down by the nozzle (though I managed to save the print! but the hole for the attachment on a shaft was ruined). It took me about 3 hours to print. Be sure to spin it in correct direction. The direction is, when viewed from the suction side, clockwise (oops! Dremel would spin it wrong!). When being spun wrongly, it still sucks (probably not as well), and creates a lot more noise.