ATX Power Supply Topology

[raptor1956]

I have a project idea that may require DC power up beyond 1200W and doing that with a linear approach will be too big, heavy and expensive so SMPS seems the way to go.  The modern ATX PS is an interesting comparison as there are commercially available models of 1200W and larger and with economy of scale the prices are not bad either.  But, I can't say I ever had the need to dig into one and don't have anything newer than about 14 years old to play with.  So, are there any pointers to the overall design and arrangement of how they're typically made? 

It looks like it will be a multi-stage unit to first of all handle incoming power of 100VAC to 240VAC and 50 or 60 Hz and correct for power factor.  The output end has multiple rails at 12VDC, -12VDC, 5VDC and 3.3VDC so there are obviously numerous sections taking an intermediate voltage and bucking it down to the required voltage.  So, is there one intermediate voltage between the filtered and corrected input section and the output sections and if so what would that be and why?

Again, in my application it would be useful to be able to leverage the design and volume production of already available ATX power supplies and if I were to contract with a supplier of them it would be easier and cheaper to use something that's off the shelf as much as possible.  If a design change is warranted and I suspect it is then having the bones of an off the shelf unit but changing the outputs would, I think, be the best way to go.

In my potential application I may be able to get away with a single output voltage in the 30VDC 25-45A range but I may need a secondary output at about 12VDC 20-30A with the total power output being, perhaps, about 1200W or maybe a bit more.  I could do a clean sheet design, but starting with a well made and efficient ATX PS and perhaps modifying the output sections may well be cheaper and quicker in the long run.


Brian

[joseph nicholas]

I'm in the process of reverse engineering a Dell ATX power supply. It uses 3 ICs to do the job.  I tried googling the data sheet but the part numbers make no sense.  Google gave me a data sheet for just one its a sc3843.   The data sheet is a clear a mud.  It seems this part is used as some kind of current limiter or cut off.  It also contains a bunch of big power transistors and a transformer, Im guessing this is there for the buck converter. 

Hope this helps.

nico

[wine+dine]

Check out the app note http://www.ti.com/lit/an/slua143/slua143.pdf that explains the 3843 nicely.

If you're looking at a 200-300W ATX supply, it is most likely a forward converter.  Look at the circuits at http://danyk.cz/s_atx_en.html, http://danyk.cz/s_atx06l.png could be close.

Note that ATX supplies have a 5V standby section, often a simple flyback circuit.  You often find something like a TOP200 series controller in these that looks like a 3-pin transistor in a TO220 case (https://www.power.com/sites/default/files/product-docs/top200-204214.pdf).

[joseph nicholas]

Thanks wine+dine.  After looking at the data sheets, it seems that there should be no problem using the uc3843 to devise a way to hack atx power supplies into usable variable voltage and current limiting power supplies.  The Dell supply I have has a daughter board which could be allow this to be desolder and a controller substuted. 

Excellent.

nico

[nctnico]

I'd go for a real 1200W power supply. It probably lasts longer at full load than a PC power supply which has been designed to supply only peak loads. Meanwell is one of the brands which has been working reliably for me and they usually aren't that expensive.

[raptor1956]

30V and 25-50A range might be economically derived by a switching post-regulator working on a standard high wattage 48V industrial / telecom PSU.  They're popular these days for computing, too.  I have seen many dirt cheap "server" PSUs that put out 48VDC at kW+ powers.

To build one from scratch you'd almost certainly use a PFC stage boosting the universal 90VAC-240VAC 50/60Hz mains voltage up to around 350-400VDC DC voltage.  Then the ~400VDC would be converted by something like a phase shifted full bridge converter to your primary wattage target 30VDC through an protectively isolated transformer and synchronous rectification stage et. al.

If you need substantial wattage (relative to the main output) on the other output DC voltages then some of them could be derived from the 400VDC with an independent phase shifted full bridge or LLC or some such converter.  But if the wattages and transformer sizing work out and your auxiliary secondary rails can be easily derived by bucking the 30VDC primary rail down to their levels then that would be a simple and low cost option.

Look at the TI UCC28950 and UCC28070 related reference designs and also the other AC to 48VDC reference designs they have.  Infineon and ON Semiconductor and IRF also have  some relevant application notes / reference designs for PC power supplies and 48VDC stuff to an extent.

You can buy COTS 48VDC PSU modules of course.  You can also buy PFC modules intended for integration in a system PSU solution if you don't want to engineer either of those parts.

Some 24VDC and 48VDC power supplies can have their outputs adjusted over a range, so you may find something applicable to 30VDC output with some simple adjustment that is supported / offered by the manufacturer.

OK now, that's exactly the kind of information I was looking for.  Thanks a bunch!

This kind of PS is not one I have much experience with and the idea of taking the 90VAC-240VAC 50/60Hz up to 350-400VDC was not exactly what I expected.  Curiously, that higher voltage and the reference to a full bridge converter almost suggest 350-400VAC and not DC and if AC probably in the tens of kHz.  I also have to confess to not being all the up on the PF stuff but presumably that would improve overall efficiency and therefore desirable particularity when running at these higher power levels of 1200W's or so.

There is much to learn here...


Brian

[raptor1956]

Once again, thanks evb149...

Actually, the biggest hurdle is exactly as you point out about knowing what information to look for and sorting through the crap-ton of other things that may not be as useful.  It is helpful when someone with experience or knowledge of the subject can focus the research and I'm grateful for your efforts in this vein. 


Brian