ATX Bench Power Supply with USB Hub

I propose re-purposing an ATX (computer) power supply unit as a bench supply with a number of USB type A power outlets (5v) in addition to +/- 12v, 3.3v and 5v for breadboards. The design will additionally feature a USB hub, an FTDI serial port and computer control.

Project Elements

atx-1.png (PNG, 5.89 KB)

Discussion (2 comments)

JohnHind 9 years ago

This appears to be the latest Intel ATX Power Supply specification (unless anyone can find a later one?):

http://cache-www.intel.com/cd/00/00/52/37/523796_523796.pdf

Checking some pictures on the internet, some of the Intel mechanical specs seem to be widely ignored, particularly the depth of the unit and the exit position of the wire harness on the rear panel. Also power inlet and switch position are not fully characterised in the spec and the aperature in the PCB may need to be larger to accomodate the variation. The mounting holes and the width and height of the mounting panel must be adhered to reasonably well or PSUs would not be exchangable in cases.

I checked several high-current (>500mA) USB 'wall warts' in my posession. The general scheme seems to be to short the D+ and D- in the A-Socket for a power-only connection delivering a nominal 2A (for example, a tablet or phone charger). This suggests some dedicated high-current A sockets in addition to some on a hub (which will be limited to 500mA).

I think we want more than four ports on the USB hub bearing in mind that one, possibly 2 ports will be taken up by control and serial port. I found this seven port chip:

http://www.ti.com/product/tusb2077a

USB-3 is tempting and gives more power flexibility, but on the whole it is probably over-kill in this application - any views on this?

Now please tell me: on what planet is 'Hi Speed' faster than 'Full Speed' - oh yes, planet Standards Committee!

0 Comment(s)

JohnHind 9 years ago

ATX computer power supplies are a tempting prospect for re-purposing for project or lab use. Mass production and extensive competition have made them excellent value and standardization to Intel's specification make them a known quantity. This idea has occurred to others before. Elektor published a breakout board designed by Ben Jordan in June 2013 and Sparkfun make a simple commercial kit.

These designs fail to hit the mark for me, and I thought I could do better.

First off, the standard mounting features of the ATX supply should allow us to make a breakout board which mounts on the supply case using simple stand-offs. This will make a self-contained unit without any mechanical construction. Sketchup models are presented to show the concept. The circuit board (in traditional green) overhangs the right-hand edge of the PSU to accommodate a standard motherboard plug (dark grey) on the back. The cable can thus be bunched up along the left side and the back of the unit tidy and out of the way. The cutout in the board allows access for the mains lead and (if you are lucky) the mains isolater switch. The board will be mounted with four bolts into the threaded holes thoughtfully provided in the ATX standard and spaced with simple stand-offs. This allows cooling airflow into the PSU which will also cool any components mounted on the back of our circuit board.

Looking around my lab bench, I see that a clear majority of my equipment takes power from a USB connector, either a cable with an 'A' connector on one end and a mini or micro USB connector on the other (or ocassionally a full sized USB 'B' connector), or a USB 'A' connector actually formed as part of the PCB. Many of these devices do not even communicate over USB but just use it as a handy source of power. This suggests that our design should have numerous USB 'A' sockets fed from the 5v. available from the PSU.

Having realised this, the next logical step was to incorporate a USB hub into the design so that at least some of these 'A' sockets could communicate as well as supply power.

With a hub present we may as well include the possibility of computer control of the unit, avoiding the need for a physical UI. Plus it would be good to throw in an FTDI serial link (3.3v or 5v logic selectable) with an Elektor ECC connector of course!

One of the problems of using a computer power supply in this way is the fact that even the lowest wattage units really produce far too much current for most lab projects bringing the possibility of serious damage and even some unintended arc welding! Some kind of current limiting is really essential. The Sparkfun design has fuses (in my lab that could get expensive!) whilst the Elektor design ... well lets just say it's not for wimps like me! I'm still thinking about this, but a simple collective 'electronic fuse' which measures current and uses the logic on/off switch to cut the supply if any output goes over limit is my current (pun not initially intended!) thinking (comments very welcome on this and and other aspects of the design).

atx-002.png (3kb)
atx-003.png (4kb)

ClemensValens 9 years ago

Just a tip: iPads and the like need voltage dividers on the USB data lines in order to enter charge mode, so maybe it would be interesting to have one USB output with variable voltages also on the data lines.

JohnHind 9 years ago

Hi Clemens, Thanks for the interesting comments (and for causing me to google "Chindogus"!). I may not have positioned the proposal well. The primary function intended is a high-capacity self-powered USB hub/charger for all the USB instrumentation, computers and development kits that accumulate on a modern development bench. It will replace multiple "wall-warts" with a single efficient PSU and provide multiple USB ports conveniently on the benchtop. The breadboard outputs are a "free bonus", not intended to replace the need for a proper variable voltage/current instrument but supplementing it by providing fixed voltages for logic etc. I never intended the breadboard to be mounted on the device, these outputs would be conventional flying leads. The grey area on the top of the PSU is a fan/ventilation aperture specified in the Intel spec. For current limiting, integrated circuits (typically 4 channels per chip) are available for USB hubs (500mA limit). For the other outputs I am thinking of an "electronic fuse" rather than a constant current limiter. It would measure the current individually on each (non-usb) output and use the logic PS_ON switch in the PSU spec to cut all the outputs if any go over-current (conveniently the spec also calls for a separate unswitched 5v VSB output which can power the control circuit). A possibility is to use Hall-Effect current sensors to avoid the problems of ground lifting or the complexity of high-side shunt sensing. As for the UI, I am a strong advocate of computer control panels (i.e. a virtual control panel on a computer or tablet). I am baffled by the continued enthusiasm for dedicated controls when the total cost of the components often exceeds the cost of even a dedicated tablet and provides an inferior user experience. With a USB hub already present, we have a handy control channel. This proposal of course also provides a convenient power source for tablets and laptops.

ClemensValens 9 years ago

The idea is of course a valid one, but projects like this often tend to be(come) Chindogus, ie devices or ideas that seem useful at first sight, but not so much when you examine them closely. Of course ATX power supplies are cheap etc, but, as you mention yourself, they are very powerfull. For safe prototyping you must add circuitry to set and limit current, which basically comes down to adding a sort of power supply as a front end to the power supply. When you have the many outputs you envision, do you protect them individually or all together? A variable voltage output would be nice too. Display? Status indication? The more features you add, the more complex and expensive the project. In the end you may be better of by desining a complete PSU from scratch. This is why we kept our ATX PSU project as simple as possible, we didn't want to throw away the price/simplicity advantage. P.S. I am not in favour of mounting a breadboard on top of the ATX PSU. It would make it uncomfortable to use, I feel. Regards, Clemens