How to begin…Okay, this is the work log for my new case project. It's called LockDown which should give you an idea of what I'm shooting for already. Here's my situation: In a few months I will have to leave the place I have called home for some 18 years and venture off to this mythical place that many refer to as "college." This is great thing for sure, but it also means that my now nearly two year old box just isn’t going to cut it anymore. One could just buy some new hardware and call it day, but I’m a bit insecure in my ability to hold the title of alpha geek once I settle into my dorm. After all, I will be attending what could be called the Mecca of geekdom, the Massachusetts Institute of Technology. I’m going to have to build a box that says: “Don’t even question the geekitude of this dude”. As with any good project, this one starts with some concept drawings. I started working on some designs over a year ago and have accumulated quite a stack of sketches of frame rails, component placement diagrams, and cooling circuit schematics, but it turned out to be all for naught. On a plane trip onboard an Embraer RJ145 from Newark NJ to Huntsville Alabama for an engineering competition at the NASA base it hit me. Vault doors are just about the coolest things in the world. The big combination locks, the awesome mechanical linkages, and the heavy bolts that extend from every side at once; I knew I had to have one. Then I started thinking, what good is a vault door without everything else being hardened too? After all, a vault door is about as big of an invitation as one can make for a bunch of geeks to try their luck at breaking in. From there the main idea became “maximum security,” plugging every conceivable hole while leaving the case easy to use for the authorized user. Make a case worthy of storing the most sensitive of information. Why? No reason in particular, really. So I threw out all of my old sketches and drew up the perfect case in about ten minutes on a seatback tray in a twenty-something seat jet bouncing up and down in severe turbulence. Throw in all the maximum security features, make it look good, and most importantly make it the perfect dorm room case. This means that it has to approach absolute silence, be small, and have plenty of horsepower. No fans, no whirring hard disks, and some heavy duty water cooling. And so here’s the main effect of the concept: Of course, the design has already changed from what’s pictured (for one, the component layout is completely different), but that’s the general idea. The gull wing doors are central to the design and will most likely be made from hardened 1/8” steel. The frame will be made from ¾” stainless steel square tube. The locking linkages are pictured on the bottom. What are not pictured are the three transmission coolers that will reside on the inside of three of the four inside door surfaces which will cool the CPU, GPU, chipset, HDDs, and the PSU. I also want to add a cast aluminum handle custom molded to my hand. This will be my third big mod job, the second from scratch, and if I’ve learned one vitally important thing it’s that Lain-Li makes a killer motherboard tray. I could try to make my own like in my last case, but it’s really an exercise in futility. Buying these couple pieces of stamped aluminum will save many hours of aggravation and let me get onto more interesting details. Before I jump into fabrication, a little more introduction is in order (like this could get any longer). I’m pretty much completely self-taught and I’ve only been doing metal fabrication as a hobby for a few years so I guarantee you will see me do many things wrong, if you know something I don’t about a particular step please speak up. This is very much a learning experience for me and will proceed very slowly. If you ever see me on the street this is what I look like: A quick intro to the shop where all this will be taking place is also needed. My dad is an auto mechanic and hot-rodder so we have a very nice shop at home. It is still a home shop though, so don’t expect to see anything too fancy happening. No CNC machines here folks. That’s not to day I don’t have a few tricks up my sleeve. This is the crown jewel of the shop, a Miller Syncrowave 250DX TIG welder. While I’m no master welder, I am pretty handy with it which opens a lot of doors with a project like this. It’ll handle everything from steel to titanium, from sheet metal to car frames. Here are the more pedestrian tools that I use constantly. 1. A steel rule – absolutely positively must have item 2. Digital calipers – indispensable for layout and dimensioning 3. Wire wheel – for cleaning and rough finishing 4. Air sander with 3M Roloc disks – for cleaning, shaping, and finishing 5. Cutoff wheel – cutting stuff in hard to reach spots 6. Mini die grinder w/ small carbide tip – forming operations 7. Random orbital sander/polisher – final finishing 8. Air drill – drilling holes 9. Aircraft snips – cutting sheet metal 10 & 11. Two full size die grinders with different burrs for rough forming And here are two more absolutely indispensable tools. The giant vise gets used for just about everything, and the handheld bandsaw is used for stock cutting and rough forming since I can’t have a full size saw. Here’s a new addition that arrived a couple days ago. It’s a very cheap Chinese knockoff Beverly Shear. If you work with sheet metal and don’t have one of these you don’t know what you’re missing, I sure didn’t. And here’s my last trick, a 3-in-1 Lathe/Mill/Drill. It’s a cheap Chinese machine and so isn’t of the highest quality, but it sure gets the job done. It’s set up milling in the picture with an extra XY table mounted just to space up the vise so it can reach mill head. I’m working on a real spacer right now. That’s it for the intros, here comes the actual worklog: I’ve been at this for a few weeks already, so this post will backtrack a bit. The first couple things on my list are rounding up the parts I need to get to build around and modifying the power supply for water cooling. Here’s the power supply, warranty voided before I could even get a picture. Here’s the motherboard tray that I’ll be working with. It is a little oversize for what I want, and I don’t need that fan mount anymore, so it got trimmed a little bit. This is the lock that’ll be the centerpiece of the front of the case. It’s a Sargent and Greenleaf combination lock, simplicity in itself but it is UL Group II certified and so gets the job done. As you can see, it’s really huge. (Those are inches, not CM!) This is the whole parts pile complete with mockup hard disks, PSU, CDROM, mobo, and the Eheim pump that’ll power the water cooling circuit. It helps me greatly to get all the parts in a big pile like this just to get a feel for how I can get everything to fit together.
Looks like I exceeded the character limit, here's the rest: The first bit of serious modification is the freshly devoid of warranty Antec NeoPower. A fan in the power supply simply won’t do in a fanless case, so this baby will get the water cooling treatment. Standard warning here: If you aren’t absolutely sure of what you are doing, don’t open a power supply. The residual charge in the capacitors is deadly unless dealt with safely. With the top popped the first order of business is to find out exactly what needs to be cooled. I loaded up the PSU with some junk hardware and turned it on. Notice the cover is off, this is extremely dangerous if done unsafely, don’t do it. I went through the PSU with my trusty infrared thermometer and found out that if I replace the aluminum heatsinks with waterblocks I should be fine. Pulling out the heatsinks is not an easy job. They must have been installed before it was all soldered up because the screws that hold it in are nearly impossible to remove. I doubt that I’ll be able to get them back in. To get a few of them I had to make a special screwdriver: It’s just a steel strip with a screwdriver head welded into one end but it goes into crevices where no normal right angle screwdriver dares to go. Now it’s time to make the waterblocks. The way I plan it is just a copper riser connected to a copper U-tube that runs at the top parallel to the PSU board. The copper I’m using for the risers is some scrap bus bar from an electrical station. I’m just transcribing the original heatsink holes and profile onto the copper with a scribe and some transfer punches. These then get cut out with the bandsaw, milled square, and then have holes drilled. The holes then get tapped which will certainly be a risky proposition with such a miniscule tap (M3x0.5) in something as gummy as copper. Our next installment will hopefully include finishing the waterblocks, but I’ll have to see how it goes.
Installment Two: Taking a picture of sawing up the copper bar wasn’t possible for me to do since I really need two hands to hold the saw so I’ll pick up with the pictures at the next step, squaring up the plates. Then the holes that I marked earlier get drilled and tapped. The tap pictured is my third one so far, the first two lie in pieces on the garage floor. Thank god they’re only a couple dollars a piece. Now the holes for the standoffs get drilled and tapped, and this is what they look like installed: Now the plates get installed into the power supply to grab some measurements for the U-Tube that will carry the water. Here it is all mocked up: With the tube all laid out I can weld it all together. I decided that the tube needs to be welded instead of soldered because I need the tube in its final form to weld the risers on. Welding on a soldered fixture would cause problems since it gets hot enough during welding to melt any solder and the whole thing would fall apart during final fabrication. The tubing joint is also included in the weld that attaches the riser to the tube and I don’t want any contamination either. Back to the riser plates, during a final checkup I noticed the larger risers interfere with the power connectors on the back of the PSU, so a notch had to be cut: With everything looking good I set it all up for the final welds that will join the risers to the tube. I can’t just make a few spot welds with everything mocked up inside the PSU because, unlike steel, by the time a liquid pool gets big enough for a spot weld the entire heatsink will be hot enough to destroy everything it’s touching. My plan here is to use some JB Weld to hold the parts together in the same position as I take them out of the PSU and spot them. Hopefully then I’ll be able to scrape out the glue and finish the entire thing up.
Your work looks awesome All electronic components generally carry charge. Did you mean the capacitors store charge?
I think he kinda means the mosfets get hot easier and 'perhaps' could use cooling. Because the caps/mofsets ofcourse store charge. Maybe this article ,from my way to large favorite list, might be of help: http://zfz.com/projects.asp?request=liquidenermax550w&page=1 But you seem to know what you're doing! Good luck!
Actually i think he means remember to insulate the mosfets from the blocks. In the pic it looks as if the blocks will be "live" with a heatsink that would be fine as it doesnt touch anything, however having a "live" waterloop would be a bit dodgy
Wow, this is cool. Has a psu waterblock ever been made before? If so, is it for sale retail somewhere?
Yes, No. You can get pre-watercooled PSUs though. Yeah, I was going to say, there's a little (usually grey) strip between the mosfets and the heatsink. That's not thermal compound (well it may help), but rather an insulator to make sure the entire heatsink isn't a live component. You NEED to keep those and put them back in place, lest have an electrified water loop (if you don't kill it entirely).
Thanks for letting me know, I was actually going to ask my "go-to" EE about that before I slapped it all back together. I figured there was something special that I would have to do seeing that the heatsinks connected to a pad on the PCB. Can you point me in the direction of some more information about this? I saved all the little gray strips, but how would they be able to insulate the mosfet with the metal screw touching both the metal piece of the mosfet and the heatsink? What exactly must be connected to the metal part of the mosfet for it to function correctly? Adding to what I mentioned above about speaking up when it looks like I'm doing something wrong, never assume I know what I'm doing. Any impression that I know what I'm doing is purely accidental.
I think only the backside of the mosfet is live. I know someone around here (or on the net, but I think at Bit) did a custom watercooled PSU. If you can find the thread, it should give you more info about the grey pads. I just know they're insulators so the heatsink [waterblock] doesn't see any power.
Love to see something modded in the way of connecting your peripherals. Maybe some sort of twist-lock connector. Did some google leg work.... Water Cooled Power supply As you can see in this article, and the sequel, he used an insulator between the mosfet and the heatsink. Some Mosfets make a connection through their plastic case to the metal tab. This is used for cooling purposes and/or as a connection signal path. In some cases the tab is used as the drain. (Voltage moves between the source and the drain, so this is where the problem comes from) So.... take note what Mosfets needed mica shims. Some shims may have been used where they weren't needed. Hope this helps!
Thanks for the help, I remember reading that article but must have glossed over that important section. It looks like no big deal, I'll be able to slip in those insulators and should be all right.
Update 5/10/05: PSU Waterblock Fabrication Continues I finally got around to welding the main structure together this afternoon. I learned a few important lessons in the process. First thing, copper is way too conductive for my taste in metalworking. To get a pool started to tack the riser plate to U-tube took about four seconds at 310 amps x ~14V. Needless to say, the JB weld tacks didn’t help in the least. About halfway into my first tack both dabs of glue disintegrated and the whole thing fell apart. I jigged the structure the best I could in my little vise and got the thing tacked and welded. The copper behaves a lot like aluminum but doesn’t have the same problems that aluminum gets with its oxide coating. Instead it gets a flaky black coating all over but it’s not nearly as much of a pain as aluminum oxide. It’s also much more pleasant to weld since AC isn’t required to blast off contaminants. Here it is before the oxide coating got cleaned off: One other thing I don’t like about working with copper is how malleable it is. After welding the piece I went to pick it off my vise but dropped it about 1m to the floor. The entire thing was mangled beyond recognition. I was pretty panicked so I didn’t get a picture, but I was able to get it back pretty close to the right shape. Here’s a picture of one spot that I polished to see how the weld bead would look: I did a few practices with welding copper plates to tube and if they are any indication the end product should be very nice. Here’s a cutaway of one of my practices: As you can see I beat the thing to hell and back to see how it would fare and then sliced it up. No cracks showed up after a few beatings and the sliced section shows near total penetration (I can pick out a couple tiny lines, whether they are gaps I can’t tell). That should do nicely for the thermal conductivity of the system. Next on the list is putting on the threaded ends for the barbs and putting the whole thing together. Hopefully I’ll be able to get out to the Home Cheapo to get the fittings soon.
I would have gone for Brazing or soldering myself, but if it works with the MIG than why not, i suppose.
