• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Redundant PSU array

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
(Hoping that this is worthy...)

Link to details

I just wanted to share what I was up to.

In short, I'm building a redundant PSU array, in an n+1 configuration, using 3 * 300W PSUs.

Yesterday, I looked into the TL494 chip's function, but I have yet to come up with a substitute to add "Remote Sense", aka V Sense+ and V Sense - .

Getting closer though...
 
it is a good idea, you will also gain redundancy..
Most swithmode powersupplys will not be harmed if they get voltage back at the outputs, so paralleling is possible, BUT
you need schematics of the PSU you want to modify for this.
Also a problem is the regulation loop, you see each psu is voltage senced and regulated at 5 and 12 volt, the 3.3 volt is often regulated separate, the negative outputs has no regulation, but they are coupled good via the output coil to give good regultaion that way. you can voltage divide the outputs and the new outputs and feed that voltahe to all PWM regulators..
I have designed switchmode PSU for many years so I know this is not easy to do, even for me, you need to understand all about how they regulate and self oscilate during powerup sequence.
else this will only end with a big-bang :)
 
I've been going over some texts, and I know what you mean.

I'm trying to add VSense+ and VSense- right now. I'm hoping that I can simply pull out a chip, and replace/patch in another one. the problem is that the +5 and +12, as you said, are regulated together, which is actually helpful (makes things simpler), but the problem is that this feedback line is essentially the equivalent of a combined VSense.

I'm aiming to use the TI chip, for redundancy. I'll post links to some great PDFs.

Looking over the TL494 chip now...

I pulled apart one of the PSUs last night, and noted the chips used.
 
oz2cpu: I just went over your webpage: excellent work! Wow!

I'm using a PSU diagram I found from a russian personal page, for reference. It is of the same DTEK 200W unit in your article, and uses the same TL494 chip I mentionned.

The Lite-on unit I selected is more complex, even more so than the 400W unit you picked apart. (looking forward to the english version:D ).

I have taken many photographs. I may post them next week.
 
Great, also I strongly suggest that you use 2 or 3 PSU of EQUAL model type, that will make things abit more simple, else you will have to reverse engineer 3 different types, it took me 2 whole days to pull the 400W type into pieces and drawing all the schematics and measuring all transformators and coils, I wanted to know all the details, even how many turns on every trafo and coil :) you dont need that detail level, but you MUST have the regulation loop and controll connections on some paper, so you know how at attack it :)
Also I suggest that you test the regulation when you are done, WITHOUT mains power applied, this is possible if you have a scope (else this mod is impossible) and a few variable lab supplys
then you can supply the controller ic, and the right voltages on the outputs !! then you see the regulation work,
you know if the input voltages are a bit too low, the controller IC will give MAX PWM out.. but if it is just a little bit too high, then it will give NO PWM at all.. at this mode, the self osc primary circuit will start to generate low output (if mains was applied)
But you have a really good advantage if you have no mains when you test and try it while working around in it.. NOTHING will explode right up your head (dont ask why I know this hehe)
and the risk to burn even small stuff are minimal..

OK REMEMBER no mains while constructing and testing :)

also Vario trafo with volt and amp meter has saved many of my constructions while powering up, then you can turn down fast if something is wrong, at an early warning state..

Sorry if I spell bad, but this was written in a hurry
and English is not my native language
 
I forgot to say good luck, and please use safety glasses :)
show us pictures, and results when done,
I am sure this is possible, with time and skill,
this is not a case for beginners,
but you sound like a skilled electronic technician or similar,
so this will work maybe not today.. but soon :)
 
Thanks for the tips! I understand everything.

I did purchase the units mentionned in the other thread (link in first post). They are all the same, 300W Lite-On units.

Yes, I did spend one year studying electronics in college.

I don't have a scope, but I definitely intend to use one, somehow, to measure the ripple. I definitely need a workspace: my wife is starting to freak out over what I put on the kitchen table :D

The loadshare controller I'm aiming to use is the TI 39002 or 39007. Link:
http://focus.ti.com/docs/prod/productfolder.jhtml?genericPartNumber=UCC39002
 
Some progress:

The modification of the PSU itself turned out to be too extensive, and complex.

Instead, I'll be using an external method, originally proposed by Yo_duh, a fellow ProCooling member.

I can't reveal any details at this point, because his article is pending publication, but I can say that putting 3 PSUs in an array is going to cost $40+.

I had to pickup 6 gauge AWD solid copper for the power bus, to cover a theoretical maximum amp load, where the 5v line of each PSU, although shared with the 3.3 line, has a max rating of 40 amps.
 
I need some advice:

I have the option of adding some circuit protection, but I need to qualify the load on the PSUs. Is a PC/mobo/drives:

-capacitive
-inductive
-resistive

or some/all of the above?
 
Yep, all parts have been ordered, and I should receive them on the 15th. Assembly will take place on the following weekend, at least in a form that is functional. I still have to tweak the PSU outputs, and I need the load resistors to do that (also ordered).
 
Back