A custom-made power distribution board (PDB) to utilize the different outputs of an ATX power supply
As a precaution:
I must remind you that i a am not a professional electronics engineer. All my make projects start with curiosity and ends with excitment whether the result was successful or not. Learning new things is the key here.
That means:
This one is such a project which is purely for demonstration purpose. So be careful when you try something similar.
Preview of my finished product
First idea
I rather make things for my home by myself instead of buying them. There is so much unused stuff stashed on shelves which I hate to throw away.
Sure, this project has nothing in common with useful home stuff. However, let’s get to the point, I had the need for a power supply delivering more than 3A…. and coincidentally, there was an unused ATX power supply laying around. So why buy a new power supply when the source is right there?
Ok, let’s make things clear, this is not about saving money. If you count hours of work, then such a project is definitely more expensive.
The goal of a customized board?
When you search the internet, let’s say for “utilizing old ATX”, then you will usually find an instruction for how to modify an ATX case by drilling holes into it, cutting out panels and wires, and doing other nasty things to your beloved ATX PSU.
There are main disadvantages to that:
As soon as the ATX PSU breaks irreparably, you have to start all over again on another one.
Also, when modifying it that way, you can no longer reuse it on a PC.
This project however is different. The key feature here is flexibility. You can plug and unplug ATX power supplies any time. You will be able to plug in ATX molex connectors directly to the custom PCB without cutting them of, but also provide terminal blocks for direct cable inputs. The power distribution board has the capability for voltage and current sensing, led indication, and more.
PCB Schematic (Final)
3D Models of my finished custom power distribution board
Tools
- Circuit Design
- KiCAD
- PCB Manufacturer And Assembly Parts
- JLCPCB
- Housing Design
- FreeCAD
- Printing the Housing
- Ender 5 Pro
- Slicing Tool: Ultimaker Cura
- Filaments: PLA (White), ASA (Black)
Preparation
I had to plan the following things in advance
- Voltages
- I want to harvest let’s say almost every power output from the ATX PSU. This means all three voltage potentials 3.3V, 5V, and up to three 12V power rails.
- Connections
- This is a key component here. Molex connectors allow plugging and unplugging the PSU at any time without cutting the connector wires.
However,, on the 3.3V and 5V power rails,, I ended up supporting only direct wire terminals. Otherwise, I had to support too many connection standards such as SCSI, HDMI, and whatnot.
- This is a key component here. Molex connectors allow plugging and unplugging the PSU at any time without cutting the connector wires.
- Size of the board
- The manufacturing costs increase rapidly as you increase the board size. More layers are more expensive as well. So I tried to keep the size moderate but set it large enough to fit in wide power traces for high currents. The PCB has a total of two layers. The lower layer will be ground for all voltage inputs and the upper layer will be positive inputs
- Sensors and other things
- Voltage, Current Sensors, Fan, and Status LED connectors. All those things must be connectable over molex. You will see at the end that it gets quite messy with so many connections densely populated on the board.
It took me a while to understand a little about high-current PCBs. It would however be far more complicated when AC had been involved, in which I had to add other parameters to consider. E.G. impedance which would cause voltage drop or ground loops causing electromagnetic noise that can interfere with other devices.
Protection
With one 5V, 3.3V, and three 12V power rails, each limited to a maximum of 10A by a fuse. A fuse is on the other hand not quite accurate with response time. So by limiting, I mean having something that is rated to 10 amps but can handle even 20 amps for a short period of time.
Clearly, a modern ATX PSU has no issues with 20A. However, we want to support a broad spectrum of ATX Power Supplies and especially the old ones which probably have served their duty long time ago.
Fuses are a good precaution as well in case you have a cheap power supply that has no protection at all.
To round it off, I added TVS diodes into the circuit to protect against overvoltage.
So in my case, i chose the following Fuse and TVS diode configuration
| Power Rail Voltage | Fuse / Rated Current | Fuse / Rated Voltage | TVS / Breakdown Voltage | TVS / Rated Power |
| 12V | 10A | 250V | 13V | 600W |
| 5V | 10A | 250V | 6V | 600W |
| 3.3V | 10A | 250V | 5V | 600W |
Mosfets
The design quality of the current limiting circuit is pretty low. It is not a dedicated design and probably doesn’t even comply with a minimum standard. However, that is good enough for my needs. I just want a little bit of control when I connect something highly power-consuming to the PSU. Even the voltage shifting over time doesn’t bother me.
However, there is one big mistake I made which I would change afterward. The MOSFET is placed on the negative side of the circuit. The issue here is that the purpose of ground protection is totally lost. It can no longer protect the user from faulty external power sources connected to the ground plug. However, as I already mentioned before, this hazard appears only when external sources are connected to the ground plug of my power distribution board. This means if something goes wrong with the integrated ATX power supply then the internal protection circuit will step in.
Soldering
I did not let the PCB manufacturer pre-solder any components which I regret later. Soldering the tiny capacitors is a pure challenge. Either you have excellent eyesight and super calm hands otherwise it turns into a mess …. like mine.
When soldering different components with different sizes or with susceptibility to higher heat, then you have to adjust the temperature of each different component. This worked quite well with the TS100 soldering iron as its iron head cannot save much heat which – to my advantage – reduces the time delay in changing temperature.
Fast Forward
We could go through all steps and each component, but this would take too much time and would overblow this post.
No, instead we say the board is finished now and we jump right to the housing assembly
I have designed many parts for the housing and some of them haven’t been used at all. I will provide you with all those models available as .stl files which you can print out with your 3D printer.
Download: ATX-To-Lab housing STL Files
Archived in: 7zip
Content: STL File
Password: atx-to-lab
The largest printable model is about 200mm in width. If your printer can’t print that big then resize all models equally by percentage. For example, making them smaller by 5%.
If you have to pre-calculate the reduced size, then here is an example
Let’s say you want to reduce the sizes of A = 200mm and B = 185mm by 5%
A = 200mm * (100% – 5%) / 100 = 190mm
B = 185mm * (100% – 5%) / 100 = 175.75mm
You can also find all these files and the PCB design files as well in my GitHub repository
https://github.com/IboschTonosch
The Base
The base of the housing is where the PDB will be fixed in its position. To hold the screws and the PCB tightly we have to put threaded inserts into the 3D-printed screw holes.
PLA material has a low melting point. That is why we can easily push a threaded insert into the slightly smaller hole by pressing it with a hot (about 250°C) soldering iron. The molten plastic will cover all the space between the teeth of the threaded insert. After cooling down the insert will hold pretty strong inside the hole.
With the screws of your choice, you can finally fasten the PCB to the housing base.
Frame
The frame has two ventilation panels on the sides to allow the air to flow from inside out. A fan will be attached later on top of the lid.
I tried to glue the panels with hot glue, but the issue is that hot glue hardens too quickly when put on PLA-printed material. There is not enough time to attach and realign any component.
Next comes the front aperture equipped with voltage displays, banana plugs, status LEDs, potentiometers, and a on-off switch. The aperture has to be equipped with threaded inserts as well to fasten it with screws to the frame. This makes it easier for later maintenance. Here, just pressing the threaded inserts with pincers was enough to hold it in place.
The assembly of the front aperture was super time-consuming and to be honest really annoying. This is clearly an assembly process that deserves extra optimization. Next time I would actually design a front aperture PCB where all the components like voltage displays and potentiometers can be directly soldered on. With that, all the annoying cable spaghetti would disappear.
Now the frame can be glued to the base after that we fasten the aperture to the frame and connect all wires to the power distribution board, as well as the power connections from the ATX power supply.
Now is the time to test the PDB. It’s the time of truth…
The Frame Lid
As this DIY device won’t be tested excessively like a real selling product, I needed some kind of visual control of what’s happening while running it. This is the reason I opted for a transparent lid. If a component begins burning or any unusual sparks appearing on the board, I would see it and could act immediately.
I know, the covered edges of the lid don’t look smooth and nice and yes it wasn’t a nice cut either. But does such a thing matter when you’re almost finished and finally can use the amazing device for other projects? Not really.
The fan is harvested from an old dead graphics card and it lights pretty fancy blue when powered. The fan is on the other hand a crucial part because the poor little heat sinks attached to the MOSFETs (mea culpa) do not provide enough mass to keep the MOSFETs cool.
Running a Load
This picture shows a load consuming about 7.84 Amperes from the 12V rail which means a total power consumption of about 94W.
Insight
There were many other issues I had to fix while building this device which I did not mention in this post. Planning and designing the project, confronting issues, searching for solutions, optimizing and so much more. Working on a full product lifecycle is such an experience. You learn so much and it is super fun. Sure this build is far from a professional solution but it doesn’t matter because anything you build up yourself is something special.
