Intro: Vacuum Chamber Cocktail Mixer
Have you ever wanted your very own robotic bartender that automatically mixes your favorite drinks when commanded to? Have you ever wanted a drink that was vacuum-infused? If your answer to one of these questions is yes, then you are in for a treat: Through this instructable I'll take you through the journey of how I made my own bartender robot that uses vacuum to pull custom drinks from the connected bottles, and serve it in units of six.
Hopefully it will show how you can cheaply utilize (near) vacuum in you own projects by using and modifying readily available products, second hand items and trash. With some simple additional mechatronics the possibilities are endless!
Now I've seen some great cocktail and bartender machines already made by other makers out there such as the lovely Bar2D2 and The Manhattan Project and this is by no means a better way to make a bar-bot, but it certainly is a different way and i hope that it as such might be inspiring to some.
Most of the mechatronic solutions to the "robot-bartender" problem seem to use either high pressure and valves to control the fluid serving or peristaltic pump, and I would highly recommend that you do the same if you are looking for a practical and rational way to make a cocktail-mixer. However this project was never meant to be rational, the final product is going to be a showpiece in our new student bar for mechanical engineering students, which means that there are basically no limits to the impracticalities I can impose on the operator as long as its cool, and I can mix in some cool/flashy mechanical and physics principals.
Its the perfect addition to any party, given that you are OK with the additional 85dB of noise it brings.
So lets get started!
Step 1: Materials and Tools
Here's an extended list of the materials and equipment that I used:
- ESP8266 - NodeMCU
Or other micro controller that you like. I'll change the ESP8266 for an ESP32 later
- Solenoid valves (x8)
You want to have valves equal to the amount of liquor and mixers that you intend to use + 1 to flush the system with air
- MOSFETs (x9)(or relays/transistors/H-Bridge, whatever you have laying about. You need one controller for each valve + one for the vacuum pump)
- Old computer power supply (Salvage)
- Prototype board (I used a small bread board as well, as I intend to change the ESP chip later)
- Wire screw-in connectors (9)
- Cable-shoes (18)
- 12v Compressor (ORVacuum pump)
- 2 Plastic sheets (60x60cm)
Due to size I would recommend getting this from a local supplier. Opaque plexiglass was the easiest for me to acquire, so that's what i used.
- 4 Threaded rods M8(1m)
- 38 M8 Nuts and washers
Out of which 8 should be self-locking
I used about 3m of 4mm hose and 1m of 6mm
- 9 Barbed tube fittings
At least 7 T-connectors for 4mm hose and two random fittings for 6mm.
- Thick bowl / Vacuum chamber
I searched second hand stores for appropriate bowls. You want something that's thick enough for you to trust it. I watched this video by The King of Random showing what it takes to implode a glass-jar vacuum chamber and the consequences, and decided to go with the beefiest bowl i could find.
- Safety Valve
Blow off valves are a bit pricey for the project, so I used a "bad" valve that i had laying around that I tested to give out at 10psi or 2/3 Vacuum
Epoxy glue (I recommend 5min)
- Plastic Weld/cement glue or Acrifix
- Silicone/Gasket material (i.e silicone or rubber)
Optional or required depending on you equipment:
- 16 Hose adapters 3D print
If your tubes and valves are not the same dimensions (which mine were not) you might need some adapters. I 3D printed mine. I would recommend making sure your valves are an appropiate size in stead as fittings can get costly.
Its nice to see the pressure in the vacuum chamber
No project is complete without some mood-lighting. I used some LED halos and strips that I found in an electronics dumpster, only the blue channel worked on the strips.
- Flow Meter
A lot of this project is assembly and fitting of existing parts and as such a general assortment of hand tools could be required such as Knives, Pliers, saws, hand, wrenches, files and sandpaper.
- Safety Goggles and protective gloves
Always wear protective googles when playing with pressure and things that can shatter!
- Rotary-tool with cutting disk and abrasive grinding head
- Hand drill with 8.5mm drill bit and a deburring tool
- Soldering Iron
- Lots of clamps
- A way to cut body/chassis
Now in this project I got to use a laser cutter at my university to cut most of the structural body parts. Although not STRICTLY necessary as most of the parts could be cut with hand tools/jig saw I would highly recommend you to lower your aspirations and change the plans to fit the tools you have available. If I were to use hand tools I would make the vacuum chamber out of existing parts (pot or similar) to make sure that I got a good airtight chamber.
I am an opportunist, by which I mean that I like to use materials that I'am able to easily and cheaply obtain around me and figure projects there after. I design this project the way it is because a friend had bought 12 solenoid valves for his brewery that he was not able to use as they did not work in gravity fed systems, and sold them to me for dirt-cheap. The only items I had to buy externally was the acrylic sheets, threaded rods, compressor and tubing. The rest was either bought second hand, salvaged or resources I already had available. I would recommend that you plan your project to some extent to fit what you already have in your arsenal as it otherwise might become a "how to save 5$ with 3000$ of equipment"-project in stock purchases such as glues and silicone.
Step 2: Plan, Prototype and Adjust
For a long time I have wanted to make my own version of a bar-bot, and the opening of our new student-bar was the perfect oppertunity to do it. Being showpiece at a bar for mechanical engineering students, I knew that the solution had to be sligtly original and more importantly: Noisy!
Recently I've been working with vacuum infusion of composites, and theres nothing more pleasing than watching the resin making its way through your dry lay-up, seemingly by magic: Vacuum is fun! One of my favorite takes on the barBot is the elegant Bar2D2drink mixer, and having acquired enough valves I knew that my solution would be something similar. I figured that a "perfect"(aka.idiotic) solution might exist in the combination of all of these; By turning the system around I could use a noisy vacuum pump to suck the liquids into the mixing chamber, and control the flow with valves.
Another couple of inspirational creations were: Convert a Tire Inflator-type Air Compressor Into a Vacuum Pump(instructables) and How to Make a Lab Vacuum Chamber(youtube).
Sketching and ordering parts
Next I made some simple sketches of how I envisioned the body and electronics of the barBot, to make out how much and what components I would need. I Decided to make the chassis in 6 "layers" of 30cm in diameter so that I would be able to fit all parts on two sheets of 60x60cm acrylics, that way it could be easily bought and transported.
Before doing final cuts in the expensive Plexiglass, I cut out the crictical structural parts in mdf and cardboard. The result was a very shaky structure. It was evident that more suport was needed, and I incorperated this before cutting the plastic parts. It is always a good idea to make prototypes with lower material investment to map out potential problems before they hit, especially when a lot of the "construction" is done on paper and computer-software.
Step 3: The Vacuum Pump
Vacuum pumps are expensive
It turns out that buyng vacuum pumps in Norway gets really expensive very fast. Wanting to keep the system at the same voltage, I started searching for 12v vacuum pumps online, and turns out its hard to find reasonable ones (even from china). As complete vacuum was not really something I need (or even want), I played with the idea of making my own and remembered back to a semi-random instructable i saw some years ago Convert a Tire Inflator-type Air Compressor Into a Vacuum Pump(instructables).This instructable shows in detail how to make a vacuum pump out of the cheap(~12$) tire-inflation compressors that many hardware and "nick-nack" stores carry. Some even report them to be able to reach 26inHg, although I've never seen mine above 22. I wont go into too much detail about the vacuum pump as Drcrash covers it well. But there is basically two steps:
It seems like most of the cheap 12v-car tire-inflation compressors are built simmilarly, in that they are just a cheap plastic shell around a small piston-compressor and dc-motor. Taking it apart should only be a matter of unscrewing a few screws. At this point you could also cut the high-pressure tubes and power cord(to a more resonable length).
Glue glue and glue
The trick to turning the compressor into a vacuum pump is covering the air inlet-hole and use the suction here to create a "vacuum". The high pressure line should be left free to. I used a piece of "T"-fitting to cover the inlet hole as it was placed quite tricky on my compressor, and I had to cover most of the top of the comrpessor to create a seal. What you want to end up with is the hose-fitting and compressor around it covered in epoxy glues, so that you know that its air-proof. So you want to glue enough small plastic pieces in place so that epoxy wont drip into the inlet hole and so forth. I cut out small pieces of plastic and used superglue to cover the surface. Then I used some "steel-epoxy" that i had, which is basically 5 minute epoxy with some thickening metal fibers in it. I mainly used it as it was thicker and easier to apply. When the glue is setting, make sure to move the part around, so that all of the epoxy doesn't pool (or even worse, drips) to one side(this is why you should probably use 5min epoxy). Once it sets, you have cheap but usable "vacuum" (suction) pump!
Step 4: Power Supply
I've found that one of the most useful items that you frequently find in electronics dumpsters is old computer power supplies(PSU = power supply unit). Computers that are broken or outdated are usually thrown away with fully functional power supplies. The nice thing about PSUs is that they provide both 12, 5 and 3.3v which are the most frequently used in mechatronics projects like these. Depending on the rating of the power supply (which is usually indicated on a sticker on the side) they are also able to provide a lot of power, which is needed for this project to run the compressor which in my case drags a whopping 4amps at 12v.
If you haven't used a computer power supply in you project before here's some tips:
Note that Wire color may differ from unit to unit, but position is always the same. So consulting a cable diagram is advised. For my unit
The green wire is the on/off switch. It needs to be wired to ground to turn the PSU ON.
Yellow wire is 12v+ (there is a 12v- as well)
Red wire is 5v+
Orange wire is 3.3v+
I needed 12v for the compressors and valves and 5v to power the ESP. I collected the wires that i needed, and gave each power cable a ground cable as well, so that they could be run separately to where they needed to be.
The green wire should be soldered to a switch along with the ground, so that you can power the unit on and off. I waited to solder on the switch until i had installed the power button in the chassis, as the switch could only be installed without wires attached.
Keep it tidy
To keep things neat I screwed the PSU appart, and cut all unused wires inside of the PSU-body. If you do this, make sure that the powered wires you cut are insulated and cant touch anything (the whole ,metal body is usually grounded and any loose leads could cause a shorting in the psu). I bundled all wires of the same color, cut them and warped them in electric tape, which should do the trick(hopefully).
Finally I used some braided wire sleeves to cover each "power + ground" pair to make it look a bit nicer and tidier. This project involves a lot of wires and tubes
Check out ProdLads instructable for more tips and tricks about turning PSUs into usable table-top power supplies.
Step 5: Controller
Arduinos are great for controlling small electronics and reading and processing small amounts of data. In this project however an ardunio (or NodeMcu(ESP8266) as I am using) just wont be able to deliver enough current or voltage to power the components. The Compressor is power hungry at 4amps and 12v, but also the valves at 0.3A and 12v is far beyond the reach of the little chip that could. So we need to extend the arduino so that it can control a higher power circuit. I find that the easiest way to do this is by "just adding" MOSFETS to the pins.
I'm not an electrical engineer: So don't follow my advice blindly when it comes to controlling the electronics. For more correct ways of controlling your BarBot you might want to search around for motor controllers etc. But what usually works just fine for me is addiing a MOSFET, so that the gate-pin of the MOSFET is connected to a digital pin on the arduino(preferably through a resistor). The source-pins are then soldered together. I use a thick gauge wire to braid the pins of all of the mosfets together and apply solder with a high temperature (~450*C), as the stp55nf06 I used were N-channel I soldered the 12v GND wire to the sources. The drains is were we now can control whether the power will go through or not, so by adding a small wire piece i soldered these to screw-connectors were the devices can be attached.
Tip: Try to keep your wires tidy, when dealing with a lot of wires and connections things become messy quite fast. I numbered all ports so as to keep track of it all.
Also: I later had problems with one of my MOSFETs, it would always be on. It turns out that I had burned it at some point ans so it had to be switched. Finding the mistake is the tricky part. Which is why you want to have a multimeter handy. I ran across the controller searching for odd voltage values and was able to find the culprit quite fast.
Step 6: Designing the Body for Laser Cutting
Laser cutting is the future for the lazy maker! I realize that each has his own preferred way to make vectors and g-codes for laser cutting, but i figured I would add some of my own experiences and tips/tricks to the mix as well.
Now the program for the laser cutter at my school does not always agree with DXFs exported from CAD-models, so I've found that creating vector files in a graphical editor offers much better control of how the vectors will turn out. For complicated models I might opt for CAD and then edit the vectors afterwards, but for the barBot I made a relatively 2D design, and it should be fine to design flat. I prefer inkscape, but again each to their own.
Scanners are great
One of the trickier parts of this build is getting the vacuum bowl to fit on top of the chamber. I chose a bowl that had an odd shape rather than round and could not just use its diameter. If you have a similar problem, scanners are great for exporting dimensions and shapes, and will (hopefully) give you a PDF with correct messurements that you can import into incskape and trace. After having traced my bowl, I cut a version in cardboard to see that it matched up with the bowl.
Work without lines
Lines add thickness to the parts in inkscape, and unless you know approximately how big the cutting line of your cutter is it will make you parts inaccurate. Either way it is a pain to work with lined objects as they will mess with Boolean operators, so avoid it. Rather fill objects with different colors and add alpha if you want to separate objects that over-lap etc.
Know the thickness of the material (and dimensions in program)
Key to deigning things in 3D in 2D is thinking about the joining and joints of plates from different directions. extra material should be added or subtracted where two plates meet each other, which is why to have a nice looking part you should know the thickness of your material, so that your joints are the same thickness, and will be flush. The correct dimensions and position of your object can be seen on the top-bar in inkscape
Align and distribute
Making accurate parts takes a long time if you don't know about alignment tools. The alignment tool allows you to center you objects together, start one object where another ends without overlap etc. This is especially useful if you want to add to or subtract from your vector with boolean tools.
What you see is not necessarily what you get. That is to say that if you don't have outlines activated, even if two squares look like one, they will still cut as two squares if they have not been connected into one vector(even if they overlap). With boolean operators you can add two overlapping or flush objects into one object with the added outline of both, subtract one from the other, get the intersection etc. This is the trick to making fast laser-cuttable files in inkscape. I like to make many small joints that are the thickness of the material. Then add them to one plate and subtract it from the other receiving plate so that they will fit together.
Exporting dxf's is not always straight forward. You'll find the inkscape-command for this under file>save copy (then choose dxf). Getting the dimensions right might take a couple of tries the first time. One of the more annoying things I've found is that in Inkscape 0.92 you somehow wont get the right dimensions unless you export in pt (not inches or mm), although this was not a problem in earlier versions.
Step 7: Constructing the Body
Hopefully if everything went alright the body should be ready for Assembly. I re-touched all of my holes with a deburring tool and a drill bit (8mm) on some of the more bothersome holes, so that the body would slide together nicely. I then measured up equal lengths on each rod so that the planes would be level, and installed the layes in the following order:
1 Bottle-holding level
I filled empty spaces in my acrylic sheet with spacers to add a bit more structural integrity. I used theese between the two layers for the "bottle-holding" layer. This is the bottom layer, so the bottom nuts are self-locking.
2 Electronics level
The electronics layer is double to deal with all the extra eight It goes on top of the Bottle holding layer, with added downward facing diagonal structures for extra support. On the top the Power supply has its own holding structure to glue it in place, on top of which the vacuum pump is glued.
With one whole swith to turn it on and of, the name might be an exageration. But it also houses most of the electronics. I added LEDs to the interior that i soldered directly to the 12v with the appropriate resistor (220ohm). This is also the right time to install the on/off button and solder it to the green wire and ground of the power supply.
4 Vacuum chamber
(see next steps)
Step 8: The Vacuum Chamber
Find vacuum top
The heart of the construction is the vacuum chamber. I searched second hand stores to find a fitting bowl to use as the clear top of the vacuum chamber. I found a few candidates, but found that most of them were eighter slightly too small to be able to fit 6 glasses within their rim, or the glass was much too thin for my comfort. I ended up using a
For the top to be able to form a tight seal, it needs to be flat. My bowl was anything but! I originally planed to sand it flat and this is what i started to do, but the bowl was way to crooked for it to be rational. Instead I cut out a smal plastic edge that followed the rim, that i glued to the bowl with epoxy. I used tape to form a "pool" around the cut out edge that i filled with glue and placed the bowl into (if that makes sense), I was stupid enough to use copper tape(since i could "shape" it) which you can still see leftovers of in the glue, I would rather recommend painters tape or duct tape. I then gave the new edge a light wet-sanding with 1200grit sandpaper to make sure it was flat.
I used a scanner to find the dimensions of the Chamber-top, and an outer and inner(slightly larger and smaller than actual rim) tracing of the top was drawn. I then measured and placed the cups inside the bowl to determine how much extra room i needed to fit the cups. The chamber was made just big enough, so as to not add to high stresses on the chamber. I made finger joints along the edges to ensure good adhesion between the walls (which would later prove to be a big mistake)
Glue, Insulate and clean
When all the parts were cut and deburred, I started gluing them together. I started by assambeling the whole chamber without the buttom. I then layed generous amounts of superglue along the edges followed by the activator. When all was glued from the inside, I put in the bottom and glued it in place. I tested it for leaks by adding water, and found many. I re-applied glue on the most noticable leaks, and then used sanitary silikone along the interior joints to fully seal it. I made sure to have a de-greasing spray handy when laying the silicone to clean of any silicone-mess right away (but use an appropriate chemical for your sealer).
When I cut out the top for the cavuum chamber i made an interior, sligly smaller, tracing of the vacuum chamber top. The outer and the inner tracing markes the start and end of where i wanted my gasket to be. By adding the inner part and applying pressure with clamps this forms a mould in which a gasket can be cast. I used dragonskin-30 rubber silicone to cast a quite soft gasket. A cheaper way is to cut it out of rubber-gasket material. Dragonskin-30 is a two component rubber-silicone. Its important to follow the mixing instructions, and use the righ amounts. To get an even gasket I used a level and hammered inn some thin splinters under the chamber to get the "mould" level before pouring the silicone. To make sure that there were no air bubbles in the silicone, and that it would pour out nice and level, I used a hand drill to which i attached a nut with some offset weights (taped on some nuts). I then ran the drill on top of the mould to vibrate out any bubbles.
Connect tubing and run wire
I connected the tubing and wire much the same way as with the vaccum pump/compressor. I superglued on a tube-connector and glued wires through a hole. I then covered it in epoxy to make it fully air-tight. The tube for the drink itself was fastened with sanitary silicone in its through-hole to allow for some flexing and movement.
On the in-line for the vacuum pump I added a valve that i measured to blow out on about 2/3 of vacuum to make sure that the camber would not under-go to high pressure.
Step 9: Reinforce the Vacuum Chamber
I was originally nervous of the glass-dome being to thin and fragile for the vacuum. But I obviously didn't spend enough time being nervous about the acrylic part of the chamber, because the first thing that happened when i tested it with high pressure (~20inHg) was loud cracking noises. Later inspection showed cracking around all of my finger joints around the edges. In retrospect, using finger joints in a vacuum chamber is a terrible idea. Every corner makes a potential for stress concentrations to build up and cracks to form. I therefor had to resort to reinforcing my vacuum chamber significantly. For the added layer I used some 6mm acrylic sheets that were laying around, making the total wall thickness 10mm on my big acrylic surfaces.
I used tape to hold the new shell together while i glued it with Acrifix.
I Mad sure to file and sand all edges and surfaces of the existing chamber flat, with slight chambers on the edges, so that it would join better with the new shell.
To bond the reinforcements to the existing body I used generous amounts of Acrifix to weld the two plastics together. Remember to use protective gloves and a lot of clamps.
The design with the build in cup holders prevented me from putting in load beams/structures on the inside at this point. Although if you are able to, this would also be a good idea.
Step 10: Connecting and Installing the Valves
Finding a way to fasten the valves were a bit tricky, as they had no flat surfaces or fastening holes. I designed the plate for the valves in two layers of acrylic plates so that the advanced shape of the valves would fit into socets, adding to their structural rigidity. Additionally I made holes on each side of the valves for zip-ties to old the valves in place while the tube adapter screwed on adding more structural strength.
Since my valves were solenoid, a fly-back diode should be installed as close to the coil as possible. I decided to simply solder them onto the connection pads
I gave each wire from the motor controller a number, and the valves numbers in a logical order as well. Then connected the ground wires with corresponding valve. The 12v + wires were nestled together and connected "randomly". I tried to do a good job when it came to cable management, but in the end I'm glad that it is a hidden part of the construction as it got rather messy.
Tubing and insulating
As the valves and tubes were of significantly different diameter i had to make adapters to reduce the tube diameter after and before the valve. I made mine based on measurements for 1/2" threads(I'm not deep into freedom units so I googled the dimensions) and my 4mm tube with barbs on the inside to grip the tube. If you are not comfortable drawing up your own adapter and you have other measurements than me this parametric file by MoonCactus on thingiverse is great. I printed the parts 102% of their original size, to compensate for shrinkage in the FDM printer. The fittings screwed on perfectly and the tubes went in snugly. In fact i was not able to find any leaks after installing. None the less I applied generous amounts of sanitary silicone on the threads and tubes before final installation.
The tubes were fitted together with T-connectors, Going from the first to the last valve and then up into the chamber. This way the whole hose length can be flushed by opening the first valve to air, while the second to last valve will be the "cleanest" mixer (aka. clear alcohol) to reduce the amount of sticky mixers left in the system.
Step 11: Geting Consistent Flow and Pouring
Making 6 cocktails at once has its own problems. One of them is pouring consistently into 6 cups at once. I had a couple of ideas during the making, but it was not until the very end that I decided on the final solution, as a lot of my trials failed. Same for all attempts is that I made a "pourer" on top of the cups, where I fed the liquid tube through the rod that held the "pourer" in place. I also ran wires through the rod for Lights on top of the pourer in the end.
Pressure guidance in cups
My initial ideal was to use tubes that went all the way into the cups, that way the cup with the most liquid in it would cause the highest pressure at the input-tube causing it to have a lower flow rate in. This however made it difficult to put cups into the machine. Also I don't know if the pressure-difference would have been large enough to distribute it evenly enough
Another idea was to just make the "spouts" level and pointing upwards, that way the same amount should over-pour at all points. This was difficult to guarantee though.
The solution i finally went for was to have "over-pressure" in the spouts, reducing the cross-section so that the pressure would distribute more evenly throughout the pouring spouts. The first prototype had way to big canals and did not pout right. I tried making some "fancy" canals and patterns for the second one by etching the pattern into one 3mm plastic sheet three times (at 100%power 80%speed 70w) and gluing it to another. This worked great, but i over-shot with the amount of glue and clogged half of the canals. For the final version I used only small amounts of super glue to old the interior structures together and rather applied more glue to the edges to insulate it. I dont have a good anwser as to how to correctly design the canals of the spout, I did it through trial and error and I think my spout is to narrow still, but it works. A more fool-proof way of doing it might be to just make the exit holes small enough, so that the total area of the six exit hole is less than the larger entry hole, but I couldn't think of a way to make that look cool, so I went with slightly random pattern which I think looks great and works well enough. Another thing to consider is that if you are using mixers with pulp, it might clog up the system here, I allowed myself to make thin patterns as my valves have filters that are finer than the pattern so I think it should be fine. But time will tell.
Step 12: Coding and Controlling the Mixer
For my project, for now any way, I decided to use Blynk. I have sort of a love/hate relationship with Blynk. It enables you to do a lot of front-end work extremely fast without having to worry about the back-end, but sooner or later i feel like my projects hit a brick wall in terms of getting the full functionality that I want. For now I want to manually control the drink machine, and think blynk will do just fine.
Simple Blynk setup
Blynk is quite straight forward, its made to be simple. Take a look at their introduction guide.
First things first would be getting the blynk app for your phone, and the library for arduino. The library comes with example sketches for most WiFi and Bluetooth arduino-enabled chips.I use the NodeMCU example. Then its only a matter of adding you wifi-name and password, and you are ready to make your app/program on your phone.
Note: NodeMCU doesnt automatically pull down its pins, I add digitalWrite(--,LOW) to all my pins in the setup so that the barBot wont turn on all valves/vacuum pump until it finds the host-server.
I added a button for each of the valves. My plan was to use a flow meter and virtual pins to have per-programed drinks. But my flow meter was not able to give accurate enough values at the flow rate the machine operates at, so for now I am using manual control and counting seconds. I've found the flow rate to be pretty accurate at 0.35l/min. Which means that one shot (4cl) takes 27seconds to pour in all 6 glasses. For more advanced programs, such as pre-programming drinks when I get e more accurate flow meter, virtual pins are used to trigger functions written on the arduino board.
Step 13: Enjoy Responsibly!
The machine has already joined a couple of pre-drinks and its been a huge hit! We are already planning version 2.0. So far its been surprisingly robust, and I haven't had any real problems with it. Except one of the valves that I think came pre-broken, so it has been disconnected.
Much like my introduction prepared you for it is not really an extremely practical machine, its slow (but steady), makes a lot of noise and requires a bit of pre- and post work. That being said it is extremely fun! And definitely a great show piece. I especial love how the spouts turned out, the pattern looks especially great with darker mixers such as grenadine.
That concludes this instructable I think: My impractical vacuum powered six-drink pouring barBot
Feel free to send me a message if you have questions!