A couple of years ago my venerable bondi iMac bit the dust. I think it fell victim of the well known flyback transformer "feature" of the earlier iMacs.

In any case, I wasn't going to simply pitch it, since at the time my research network consisted of three 6100/60's, all running mkLinux. 233Mhz was definitely speedy in comparison.

Fortunately, a large body of work on how to rewire an ATX powersupply to feed the iMac motherboard already existed on the net. However, I didn't want just another beige box sitting on the ground, I wanted something I could rackmount, or at least look "rackmountable", and in black, preferably.

I looked at marathon computers "iRack" solution, but since I was (and still am basically) a poor college student, ponying up $400 bucks for their kit was not really an option.

So, I spent a couple of hours with DesignWorkshop Lite, and built up a roughly 1U case that would contain all the necessary bits and pieces of the iMac. I've since lost those plans, sorry.

I settled on 1/4" hardboard as the main construction material, basically due to its workability, price, and ability to take paint. I've worked a lot with the stuff before, and although my father used to deride it as "cardboard", everything I've ever built with it has stood the test of time. Just be sure to paint it before getting it wet.

Anyway, the case is 18" wide by 16" deep, and 2 1/2" tall, which makes it larger than 1U, but still pretty small. Construction is straightforward, and if anyone wants the design plans, I have them on a backup CD somewhere.

I used a radial-arm saw to cut the top and bottom pieces, as well as ripping the sidestock. I used a scroll saw to cut out the various openings, and then dressed them with a dremel tool. I used a router bit chucked into a drill press for some of the vertical routing that had to be done, like the CD/HD mounting slots. None of this stuff is really cabinetry-level work, and since I had all the pieces already laid out virtually, it only took a day to bang out all the pieces. The wood was glued together with gorilla glue and TiteBond II aliphatic resin. For those without access to nice woodworking materials, I've since done similar work with one of those rotary cutters (think RotoZip), but it's a lot more hassle.

For the power supply, I bought a real cheap ATX powersupply off of pricewatch.com for $8, and removed the metal casing and fan. I got a real cheap one because many of the more expensive power supplies have heavier-duty heatsinks and transistors, and I couldn't have anything sticking up very high, or it wouldn't fit in the little pizza box I was planning. As it was, I did have to hack off the top 1/2" of the existing heatsinks.

I pitched the trayload CD drive and replaced it with a cool AOpen slot load that I had laying around. In case the CD drive ever croaked, I cut out most of the front panel, and instead had a second faceplate piece that I could screw to the real backplane, and that would allow me to simply replace that one piece if I ever had to. I also installed the power switch, and four status LEDs in the faux face.

A note about the LEDs:

The green LED I simply fed off of one of the unused 5V power lines coming from the power supply (plus an appropriate resistor, of course). This means it's on whenever the computer is on. The other three I had planned for network or HD activity. Unfortunately, the iMac doesn't have any network status LEDs, or does it?

If you look closely at the iMac motherboard, near the "CUDA reset" tactile switch, you will notice that there are indeed solder pads and silkscreening on the mobo for LED positions, although there are no LEDs or accompanying resistors actually present.

Taking a bet, I figured that the values for the LED resistors would be the same for the power supply status LED that is very near the ethernet LEDs, or about 330 Ohms. Once I hooked up some LEDs with the correct resistors, and soldered their leads to the SMT LED pads, they worked great. I did notice however, that contrary to what I would have thought would be common practice, the pads are not brought to +5V, but instead brought low. So, I tapped into the backplane voltage via another unused spot on the back, one of the "FS3" SMT pads.

For now, I left the HD activity lead unconnected, because my future plans for the box were primarily non HD-intensive. The power switch on the front panel simply switches the 110 coming in.

Everything else was pretty straightforward. I decided to use the original iMac filter board, with it's huge filter inductor, more because it had the power supply cord connector on it than anything else. To move power from the filter board to the mobo I had to clip off the pins, and then reconnect them with a bundle of wires. The connector I clipped off the filter board I then soldered to a bit of perfboard, and ran the wires to that. It worked out quite well, and just barely fit into the space necessary. Notice that there is a quite a bit of dust, that's due to this box being in operation for quite some time before I got around to taking the pictures.

The mobo itself is suspended from the bottom of the box by several standoffs. If you have a modem installed, like I did, that pretty much dictates the height the mobo is going to sit at, because it extends from the bottom of the mobo.

To mount the HD and CD drive, I used my previously mentioned makeshift vertical mill to mill slots in some 4" long pieces of wallstock. These were then glued onto the floorplate of the case. I screwed the appropriate size screws to the sides of the CD and HD in the mounting holes, but installed spacers inbetween the screwhead and CD/HD case. These bushings then fit down inside the milled slots in the mounting brackets.

I could tell that adequate cooling was going to be a big problem in this thing, because of the power supply (with it's short heatsink fins) and the CPU heatsink. I always laugh at the sterotypical pc modder, and his case crammed full of fans. Well, I didn't have the room nor the inclination to have a bunch of small ones, so instead I decided on one big fan.

I found a surplus "muffin" 110V fan that looked like it would work. Since I needed to be able to lift the top off of the case to work on it, I decided to build the ducting into the top of the lid. If you look at the (very dusty) surround, you can see that the fan is covered on three sides, with the fourth (left) side free to blow air across the power supply's heatsinks. The other duct runs to blow air over the CPU heatsink on the motherboard. I added a bit of foam padding around the inside of the fan surround so as to make an air-tight seal around the top of the fan (the pink stuff). Notice how dusty it becomes.

Other engineering troubles involved how to get the AV board of the iMac in there. I could have simply tracked the traces on the PCB, and wired the video out of the motherboard into a VGA connector, but that would have necessitated my rewiring the IR and sound as well. Although I did end up installing the IR module (you can see it to the left of the power supply heatsinks in the pictures), I've never used it, and there isn't even an opening in the faceplate for it to "see" out of. I've never used the audio out either.

The hardest part of the project was correctly wiring the DB15 (old macintosh style) monitor connector to a standard VGA monitor connector. I found three wiring diagrams on the 'net, all different, and all wrong. I eventually just used a multimeter and a commercial mac-to-vga adaptor and found which pins connected to which.

Another problem was the proximity of the hard drive connector to the motherboard IDE connector. A standard ribbon cable had to be pretty convoluted to make the twists and turns necessary, and that area was already crammed full of video and power cables. To fix this, I bought a round IDE cable, and cut the sheath off of it, leaving the free wires to bend as they please. Although it looks like a mess, and probably screws with RF interference dampening, I haven't had a bit of problem with it. One other thing: The IDE connector on the iMac has a full 40 pin header, whereas many IDE cables have a pin missing. This makes plugging a normal IDE cable into the iMac's connector impossible. Many cheaper IDE cables don't have this, but I needed my sheathless round cable to work, so I clipped off the offending pin inside of the iMac's connector. It's not used for anything within the IDE spec anyway, so nothing was hurt by me doing so.

A few more things that I needed to get around was how the iMac signalled the ATX powersupply to turn on. ATX powersupplies need a low (ground) signal to turn on, whereas the iMac uses a high (+5V) as a turn on signal. To fix this, I simply used a NPN transistor and wired it up as per instructions I found on the net, although it's pretty trivial to find out how otherwise.

Lastly, I primed the case with automotive primer, and then covered the case with Rustoleum "hammered" finish paint. It gave the machine a nice "anodized" look and converted the wood to a very metallic feel.

Ever since I've built this machine, it has been happily grinding away as a DNS server in my network. The only drawback to the design is that I have to take the lid off and blow out all the dust that accumulates. This is to expected due to the massive amounts of airflow the fan creates, and all the nooks and crannies the dust can find. If I don't clean it out every six months or so, it can get a mite unstable because of high CPU heat, I think. Other than that, it's as rock steady as they come. The fancy blinking lights are a nice feature too.

iMac Rev. B to ATX conversion: link.

Note: the Rev. A and Rev. B iMac mobos are exactly the same, except that Rev. B mobos do not have IR or mezzanine slots.

iMac power on signal conversion: link.

PriceWatch.com for cheap (read: not expensive) computer components.

Jameco.com for electrical components. Don't feed the RatShack pigs, if at all possible.

Marathon Computer iRack iMac conversion kit.