Posted on

Hoverlay II

aah it looks gooood :)

Hoverlay is still a floating-in-mid-air rear projection screen just like the Hoverlay I.

Check out the video to see it in action:

A System Design Document can be downloaded here

Preliminary Assembly Instructions

Related the Project

It’s main advantages over the Hoverlay I:

  • it yields way better image quality, count pixels if you want
  • it finally IS stackable, put together as many as you like for getting a bigger screen ( many as your power supply likes.. actually)
  • it’s completely open source / open hardware under GPL V3 license, so you can build your own using the files in the GitHub repo.
  • it features some interactive visuals and applications where you can actually interact with objects on the screen
  • it has a cyber glove ^^ oh yes

While it’s only two purposes are (sorry if this might disappoint you):

  • looking as fancy as possible
  • giving me a hopefully good grade in my video technology project course.

And the disadvantages over the Hoverlay I

  • has no dedicated article on hackaday (yet) 🙂 thanks Brian!

By the way, the technical term for this kind of device is vapor screen or fog screen.

It wouldn’t be a worthy successor of the Hoverlay I if it wouldn’t come with a fancy info graphic.
Hoverlay II
The first functional build of Hoverlay II letting the twitter bird fly.


Lessons learned

As a start, I want to throw in the lessons learned from the issues Hoverlay I, or as I actually ambitiously call it the version 0.1, and how I approached to resolve those issues.

Learned #1 – Watertight

Hoverlay I was leaking like a toddler. Not because the vat had a leakage, it was because the fog just went everywhere in the housing, even against the direction of the streaming air, eventually condensing on the inner walls and wetting the fans.

Hoverlay I: Forking the air stream into the fog channel lets water and fog leak into the air channel and onto the fans and ground.

The Hoverlay II has fully seperated channels for air and fog, so that no fog could ever reach the air channel with the fans. Also, condensed fog from the inner walls would be guided back into the vat by the walls of the vent, which are now designed in a way so that they are sliding in the inner part of the the vat and not the outer.

Hoverlay II: No forking, no leaks. The walls of the fog channel guide condensed water directly back into the vat.

This is made possible by the Venturi inspired fog channel. It’s called “Strömung nach Bernoulli und Venturi” in German while in English it’s just the Venturi Effect, and it describes the dependency between velocity and pressure of gasses. Basically it states, that whenever a gas is moving fast, this is always accompanied by a preasure drop within the moving gas relative to a still gas. In the case of the Hoverlay II, the rapidly flowing air after the air escape carries a relative underpressure compared to the stil air-fog mix after the fog escape. This pressure difference simply sucks the fog out of the fog channel. Hoverlay I probably also made use of those effects, but I didn’t recognize that I could use the Venturi Effect to seperate the “dry” air channel from the “wet” fog channel completely, preventing condensed fog and therefore leaking water. In the Hoverlay I, I forked some of the moving air into the fog channel to lift the fog out of the vat and feed it between the flow formers. In Hoverlay II, the Venturi Effect, that causes a preasure drop between the outputs of the two flow formers, sucks out the fog from the center fog channel. A perforated PVC tube allows pressure compensation in the fog channel.

Learned #2 – Crisp Image

In the Hoverlay I, fog came out through a 20 mm wide slit between the two parallel air stream formers. The image turned out to be quite blurry, and went even more blurry 50 cm above the slit. I figured the blurryness was due to a too large fog-layer thickness. By experimenting I found out that reducing the width of the slit in conjunction with converging air stream formers helps to keep the fog-layer width narrow enough to get a clear picture even 50 cm above the slit. I could not really capture the difference in image sharpness between Hoverlay I and Hoverlay II yet, but if you see them next to each other it’s extremely obvious.

Learned #3 – Simplicity

Hoverlay I ate up more than double the material of the Hoverlay II and was difficult to put together. This is mainly due to the box-like shape which I still consider as a pleasant design principle for furniture-like products. However, it create necessity for the functional walls on the inside and the housing walls on the outside, thus requiring way more material than actually necessary to do the job. With Hoverlay II, I took the principles of form-follows-function and multi-purpose-parts and therefore was able to reduce the total amount of plastic by 50% and the number of plastic parts by 50%, too, which excludes those tons of straws in the stream former. There are also way less glued connections, so that the device is way easier to put together.

Learned #4 – Chainable Devices

I didn’t document this very frustrating aspect of the Hoverlay I in the blog, but finding a good way to make the devices chainable, such as “just plugging two Hoverlays together to have a bigger screen” was not as easy as it sounded like. It would require a very strong electrical connection that I wanted to seamlessly integrate into the housing. With each hoverlay consuming about 100 W of electricity at voltages between 12 and 24 V, a chain of hoverlays could easily have about 500 W, with about 15 A running through the supply wires. That also means, that the first Hoverlay in a chain of five would have to be able to pass through 12 A though a electrical connection to the rest of the chain.

Concept Art-01
Quite a task: The seamless and easy-snap electrical conection between those Hoverlay units in an array of five Hoverlays has to be able to take about 15 Amps.

My first try was using two cylindrical Neodymium magnets with a center hole, which should work as a connector and were easily attachable to the housing using a screw. Well, it didn’t work at all, the magnets faces weren’t flat enough to provide a large enough contact area and thus quickly heated up on the added resistance. Also, they add additional space between the boxes. It would have been nice if it would have worked, though, because magnets are quite inexpensive compared to those nice gold plated connectors (and even compared to regular banana connectors) and snap together the convenient “MagSafe”-way. Eventually, I found some gold plated connectors that were cheap enough to replace the magnets. They didn’t have any mount like typical banana connectors for project-box installation, which was good, because those mounts usually stand out of the housing a few millimeters, not allowing semless snap-together like required here. 3D printing a custom adaptor plate did the trick.

12 of these little guys hold the connectors in every hoverlay module in place
2014-06-06 17.45.06
those gold plated connectors are held in place by the above adapter plate and seamlessly interconnect the hoverlay modules

Still, the max. current rating of those connectors will always set a limit on how many hoverlays you can snap together. For now, 5 should not be exceeded.

Other Improvements

3D Printed Flow Formers

As in the Hoverlay I article described, the technology requires a flow forming device I used to build from large arrays of straws I cut to length and melted them together. That worked great with parallel straws in parallel channels. For saving material and milling time I replaced this air channel, that formerly consisted of many plastic walls with an air channel, that only consisted of two walls. Since those walls aren’t parallel anymore, I had to replace the straw array with something else that can accommodate convergent walls. I decided to give 3D printing a try, which was not so easy, because this would require to print very thin walls. After some first test I noticed that more difficult than printing thin walls is obtaining useable results from passing 3D models through the slicing software. I printed several iterations of different array designs I generated with OpenSCAD which I passed through the normal toolchain with no satisfying results. Also, sending rather complex objects with many triangle faces and many thin walls throug the slicer is a time consuming job, most of the slicing jobs took over 7 hours, and printing one of the resulting gcodes another 13 hours. I tried CURA and Slic3r with many different configuration, both generated very different results, but neither software gave me a good one.

2014-06-06 17.45.25
little hint: in the end the flow formers were printed successfully

For what seemed to be a quite simple task, I already had wasted many hours of redesigning and many hours of slicing. I then found the solution by dropping the regular toolchain and writing a short Processing sketch that generates the gcode for the printer directly. It would only have to lay down the material in a simple repeating honeycomb pattern which I already had written in openSCAD, so I didn’t even have to rewrite the parts that generate the geometry. I was lucky, the first print of the self generated code was very promising, I just had to get rid of some little ooze in the air channels. Optimizing the travel moves so that the printer head would not travel directly between points, but only travel on lines it previously printed, fixed that. The Processing sketch took 30 seconds to generate the gcode and reduced the printing time and material consumtion by 50%.

Bigger Water Tank

The water tank of Hoverlay is reasonably larger than the one of the V1. It can hold almost double the amount of water for longer runtime without refill and up to 7 atomizers if necessary.

Cool Logo

Hoverlay II also got a logo on it, since – I found – without, it just looks too plain. The logo derived from the side view of the Hoverlay and shows the two air channels. If the other improvements didn’t catch your attention, I hope this will.

Further Challenges

Power line feedback of parallel driven atomizers

..or whatever an expert might call it. For the reasons of component availability and for having a combined power supply for 24 V AC and 12 V DC, the hoverlay is driven by two powerful (500 W) 12 V AC transformers in series. Two bridge rectifiers together with some capacitors are used to get the 12 V AC required by the two symmetrical rows (front and back) of 12 DC fans while the series connection of transformers still yields 24 V suitable for the atomizers, which are all connected in parallel. I am quite sure that it is impossible to drive 10 atomizers in series from 230 V ~ mains, and I also didn’t want to have mains voltage in the vat. This setup is rather bulky though, which is why I installed in a Mini-ITX housing. A more elegant solution to the power issue is still on the to do list. Because these atomizers are not intended to be used together in one power line, this setup initially suffered from high frequency feedback of the atomizers back into the power supply line, randomly causing atomizers to fail in a spectacular little under-water firestorm. I was able to fix the issue simply by adding a typical mains filter rated 230 V AC 16 A right into the secondary circuit of the transformer before the rectifiers.

Interactive Applications

The Hoverlay II has been equipped with several interactive applications that involve a Kinect sensor, music and, yes, a cyber glove. All applications have been written using Processing and will be available in the GitHub repository very soon!

The cyber glove features a Sparkfun Pro Micro tied to a NRF24L01+ module and a pager motor driven by a logic level BSS98 MOSFET. Other components: A L78L33 voltage regulator, some resistors, some caps, a battery clip. It’s adressed by an Arduino Uno with a NRF24L01+ backpack.

Super String:

This application enables you to draw your own sound activated visuals on the hoverlay. The Hoverlay screen forms a multi touch screen on which you can draw colorful 3D shapes (“strings”) with your hands. After a shape has been drawn, it will start to “live” by deforming with the beat of music running in the background, as well as moving and rotating in 3D space. When the cyber glove is used, proximity to the touch screen is indicated by the glove vibrating softly in the rhythm of the music.

This video gives you an impression of how it works:


​This app shows a small satellite floating in outer space that can be interacted with by hand, such as turning it arround, poking it. When the glove is used, proximity to the screen is indicated by a soft vibration in the glove. The satellite was taken from the objloader processing library examples.

Cyber Pong:

Two players play a classical pong game, controlling their paddles by moving their hand up and down. The ball gets faster over time and a small bonus apple has been implemented: If a player hits the bonus apple with the ball, his paddle’s size grows to double size for 30 seconds. And of course it has some sound effects.

MSAFluid Canvas

One of the first ideas for an interactive application on the Hoverlay was to have some particle and plasma effects on it that can be created and manipulated by touching the screen with your hands. The MSAFluid Canvas application, as the name already tells, makes use of the amazing MSAFluid library to achieve exactly that. When the glove is used, proximity to the screen is indicated by a soft vibration in the glove.


BOM (work in progress)

All filenames are according to the e GitHub repo, where you can find all the CAD data for the DIYable parts.

Milled parts (for lasercuting: add drillholes), the material is 3mm PVC or acrylic (never put PVC into a lasercutter)

  • 1x becken-boden.dxf
  • 1x becken-seite-2x.dxf (contains two parts)
  • 1x becken-stirn-2x.dxf (contains two parts)
  • 1x kamin-2x.dxf (contains two parts)
  • 1x lufteinlass-2x.dxf (contains two parts)
  • 1x m-seite-logo-neu.dxf (logo is for engraving, not a cut-out)
  • 1x m-seite-schriftzug.dxf (text is for engraving, not a cut-out)
  • 2x m-stirn.dxf
  • 1x nebler-halter.dxf (contains 3 parts)

Printed parts

  • 12x stecker_halter.stl
  • 1x stromversorgung_1.stl
  • 1x stromversorgung_2.stl
  • 4x supercomb_air.gcode (generated by supercomb_air.pde, a processing sketch)
  • 2x supercomb_fog.gcode (generated by supercomb_fog.pde, a processing sketch)

Other Materials

  • 4x ultrasonic atomizers
  • 1x 326mm long piece of ø25mm PVC tube (perforated with a driller)
  • 8x 80mm PC case fans (as much cfm as possible)
  • 6x male 3,5 mm rc car battery power connectors for the Hoverlay II (i got these)
  • 6x female 3,5 mm rc car battery power connectors for the Hoverlay II (i got these)
  • 6x female 3,5 mm rc car battery power connectors for the power supply connectors
  • 12x cable ring shoes (solder them to the connectors for attaching them to the stecker_halter.stl parts
  • strong power cables, more cable shoes, solder
  • plastics glue (for putting together the housing)
  • double sided adhesive tape (for installing the supercombs)
  • silicone (for water-tightening the vat, water tightening solder connections)
  • various screws, length mostly dependend on your case-fan height
  • appropriate power supplies for your electronics
  • a video or still image projector. A 3000 ANSI lumen projector doesn’t hurt while a overhead or dia projector will also work. LED projectors are not recommended, for holographic projections use a light-field projector and/or add a few drops of LSD to the vat*

(*sryl, never do that!)

Build Instructions

Not yet available, sorry! Use the info graphic 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *