Open Source Power Mac G4 Cube VRM

[evanboonie]

Repo Link:​

GitHub - evanboonie/CubeVRM: High power VRM for the Power Mac G4 Cube

High power VRM for the Power Mac G4 Cube. Contribute to evanboonie/CubeVRM development by creating an account on GitHub.

github.com

Details:​

After performing some upgrades to the stock VRM and further discussion on designing a new VRM in my thread on that topic, I now have a schematic for a completely new VRM design. This new VRM is being built with two primary purposes in mind:

1. Allow larger video cards to be installed in the cube without relocating the VRM. This will be accomplished by placing all components on the opposite side of the board from the 30 pin VRM connector.
2. Support running dual, overclocked 7448 processors and a Radeon 9800 Pro video card.

Specs:​

• 21A on the 3.3v rail (all on 30 pin connector)
• 23.5A on the 5v rail (15A on 30 pin, 8.5A on Molex)
• 11.5A on the 12v rail (3A on 30 pin, 8.5A on Molex)

The Over Current Protection will not kick in until each of these limits are breached by a few amps. The ratings above are the max ratings for the connectors themselves, and they may heat up, melt, and/or burn when pushed too far beyond this limit. The OCP should prevent any damage to the VRM if it is overloaded.

This VRM is currently in an early state of development and is not ready for manufacture or use. I will continue to update this thread as progress is made.

 

[EmilioCube]

Wow! That sounds awsome! I am currently messing around with my VRM that I wanted to solder the 4840B FETs on but two pads just keep connecting themselves causing a shortcut. In case I fail soldering the FETs on the VRM, I will definitely buy yours when it is ready for “Production“!

 

[evanboonie]

Wow! That sounds awsome! I am currently messing around with my VRM that I wanted to solder the 4840B FETs on but two pads just keep connecting themselves causing a shortcut.

When hand soldering small SOIC parts like this, I find that it helps to use a very fine point soldering tip. Wet the tip with an extremely small amount of solder, then come in parallel to the pins one at a time. I try to avoid adding solder while the iron is touching the pin, only adding solder to the iron alone and bringing it in as needed. If you get too much on there, just suck it out with desoldering braid and try again. I have never had much luck drag soldering smaller components like these FETs. Adding some flux after everything is tacked down and touching the joints with the (very clean) iron again usually helps clear the smaller bridges that can occur.

 

[EmilioCube]

Thank you for your reply. I actually desoldered the old six 4840s with a hot air station and then gave my best getting the old solder away with an desoldering wire but when pressing the soldering iron and the wire on the board, the green “isolation” between these two pins was gone because I pressed to hard. Under the “isolation” there is something looking like copper that my soldering paste of course is also connecting to. I will maybe come up with a picture of that when I have the time today. So I applied the soldering paste, heated it up and everything behaved as I expected except for the thing that these two pads behaved as they would be one big. I got all the other seven 4840Bs soldered onto the board with the hot air station without problems but even after trying my best with getting the solder away from these two pads with a desoldering pump and the desoldering wire, when applying some solder separately with a thin soldering iron, the problem still occurs and the two pads keep connecting.

 

[ToniCH]

I will follow this closely. 👍

If it ever comes to the stage where I could order a parts kit and a circuit board from somewhere to solder it together myself I would be very interested in it! My Cube is just waiting for such project but I cannot design it myself so I must rely on people who do have such skills. I have been eyeing the ready made Japanese VRM but I consider it too expensive for me.

 

[evanboonie]

Thank you for your reply. I actually desoldered the old six 4840s with a hot air station and then gave my best getting the old solder away with an desoldering wire but when pressing the soldering iron and the wire on the board, the green “isolation” between these two pins was gone because I pressed to hard. Under the “isolation” there is something looking like copper that my soldering paste of course is also connecting to. I will maybe come up with a picture of that when I have the time today. So I applied the soldering paste, heated it up and everything behaved as I expected except for the thing that these two pads behaved as they would be one big. I got all the other seven 4840Bs soldered onto the board with the hot air station without problems but even after trying my best with getting the solder away from these two pads with a desoldering pump and the desoldering wire, when applying some solder separately with a thin soldering iron, the problem still occurs and the two pads keep connecting.

Ah, I see. Tearing away mask can definitely make soldering difficult, but there is a high chance that it is okay in this case. Pins 1-3 and 5-8 are internally tied together on these FETs. So as long as it isn't pins 3-4 that are bridging, it will still work fine.

 

[EmilioCube]

So I tried to solder the bridge away and it worked, do not ask me why it did not work earlier. But when I finished soldering the FETs back on and attached some small heatsinks, built the VRM back in the cube, made sure everything in my cube was fine and powered it on, I saw the white light from the on/off-button and thought that I made it. But just two seconds after the white light from the power button, the green light on the VRM and the red light on the motherboard appeared, there was a loud short circuit in my Cube, something (small) exploded and a flash appeared, and I instantly plugged the power cable off. 5 seconds later it begun to stink a bit.

What was that? Is my motherboard now dead? In a thread I read a long time ago, somebody said that I have to check the resistance on specific pins on the FETs on the VRM. Should I do that?
After that appeared, I read your reply and that should not be the issue, because the two pins were 7 and 8 either way.
Besides from that, I made no changes to my cube, so the problem has to be the VRM.
Please help!

 

[evanboonie]

So I tried to solder the bridge away and it worked, do not ask me why it did not work earlier. But when I finished soldering the FETs back on and attached some small heatsinks, built the VRM back in the cube, made sure everything in my cube was fine and powered it on, I saw the white light from the on/off-button and thought that I made it. But just two seconds after the white light from the power button, the green light on the VRM and the red light on the motherboard appeared, there was a loud short circuit in my Cube, something (small) exploded and a flash appeared, and I instantly plugged the power cable off. 5 seconds later it begun to stink a bit.

What was that? Is my motherboard now dead? In a thread I read a long time ago, somebody said that I have to check the resistance on specific pins on the FETs on the VRM. Should I do that?
After that appeared, I read your reply and that should not be the issue, because the two pins were 7 and 8 either way.
Besides from that, I made no changes to my cube, so the problem has to be the VRM.
Please help!

I'm taking this over to the upgrade thread to keep this one on topic for the new design.

 

[EmilioCube]

.

 

[EmilioCube]

Since my VRM is trash now and there is no stratos VRM available in the internet i will have to wait till your design is done.

 

[lolof]

Very nice. Here the VRM on my cubes are still running. I have 3...
One of my cubes has a sonneth 1.2 Ghz... If you get it working and finish, I would be happy to build at least one for my 1.2Ghz cube.
I wish you to have sucess with it. I am sure we are a lot to follow you on your build.
I am considering to do the upgrade one VRM as you described on the other thread.

 

[EmilioCube]

Has your design and production process came any further in the last six and a half months?