Still in progress.

I have to review my notes here, and get this to move along again. One of the tested circuits had a faulty MOSFET, because I damaged it when I built it (bent pin), but that's ok, I have a spare.

Otherwise, I've determined that one of the three PSUs is going to have to be fitted in the lower area of the case, because I just can't do with only one 5 1/4 bay. That's going to involve a little custom work, but it should be pretty easy.

I'll post pictures of the progress, as it comes. Right now I'm working 2 jobs, so I have little spare time :( . My website is requiring a little maintenance, and I'm in the process of purchasing some test equipment. Busy, busy, busy... I want to get this PC up and running within a couple of months though (fingers crossed).

I must say this was an interesting read. I read all the posts on all the pages. Well...some of the real technical bits I just skimmed over ;)

Good job BB2K with all the documentation.

I am amazed you have stuck with this project this long. We all have projects that start off strong but die out slowly as our interest in them drops over time. Also, having made a few fanbus and LM317 based controllers I know how the cost of all the parts adds up fast.

Thanks!

Yeah, the cost exceeded the purpose of the project: dodging the outrageous price of a fat PSU :D but I'll have redundancy.

I'm shopping for an oscilloscope, so I can complete this.

Now if I could only lay this all out somewhere so that the wife doesn't complain...

http://www.silverpcs.com/product/PS_PCPPT60ATX

Can this be considered a true 500watts PSU? Maybe 600watts since the peak is 700watts.

Is this thread dead Ben? I came across something interesting and I suddenly remembered seeing this thread. Any use to you?

Interesting, but I'd need more than one of those.

The project isn't dead. Unfortunately, I'm very close to a move, and spent most of my time on the water cooling testbench, and my site updates. I've squeezed some extra time to finish my case mods, and I'm about to pull out everything I've done so far, to finish this project; kinda silly to hold on to a mobo, CPU, block, pump, hdd, and all that copper...

One of the problems with using diodes is that they cause a voltage drop (some diodes it is higher than others) which is generally undesirable.

Also, due to the large quantities of current running through your typical psu, it would require quite a hefty diode (it also would more than likely get pretty hot).

Sorry that the mosfet solution did not work in practicality bigben2k, too bad that working in theory is not good enough :rolleyes:

It ain't dead yet!

Now that I have an oscilloscope, and a good DMM, I can get some serious work done on this.

Just you wait... ;)

At long last, this project is revived!

the front plate of the PSUs (while the grill is being cut out):
http://wbta.us/bigben2k/PSUfrontplate.JPG

Then the front plate mounted to a 5 1/4 tray (ripped out from dead CD-ROM drives) and a blue LED fan:
http://wbta.us/bigben2k/PSUwithfan.JPG

Then the same unit with the PSU PCB mounted, two 6" ccfls, and the handles:
http://wbta.us/bigben2k/PSUwithCCFLs.JPG

and finally, three of the above, fitted in the case (minus the handles):
http://wbta.us/bigben2k/PSUarraymounted.JPG


Much more work to do.

Re: 300w + 300w = 600w?
 

Thought i'd let you know i found the thread, read it all, and gained a blazing headache ;- )

Not so bad actually. Im allreay convinced that this is far more advanced then what i thought it should be, and still, i havent understood half of it.Seeing the time, and money you spent on this, im starting to think about looking for an actuall hotpluggable, redundant psuarray, and reverse engineer the power distributor. Heck, it might be possible to find a broken down one, and go from there.

Thanks for tipping me of on this thread, and i have to say im sorry you never found a good solution. Or so i assume, seeing as it never got completed.

Best wishes.
B!

Re: 300w + 300w = 600w?
 

I believe that you'll find that this whole thread is essentially split in two; the first part has info on a failed effort to integrate a VSense (+ and -) into an existing supply; you'd have to be an engineer to do that, I concluded, and beyond me.

If you look into ready-made redundant supplies, the first part of this thread will be relevant.

Then Yo-Duh_87 came up with a circuit that sits externally to the supplies. That's the second part.

The circuit is simple, in that it's designed for a single voltage rail. It can be applied for the other (positive) rails, but I haven't tested it. A component swap would be required for the negative rails, if so desired, but I opted to use the first supplies' negative voltages rails only; the negative rails work off of a simple regulator, and there is no requirement for additional power. It's a bit limiting on the redundancy aspect, but 3 PSUs would be required regardless (for N+1, using 300W supplies), so pulling out one of the other two still gives you "hot-swap" capabilities.

You might consider running the negative voltages off of a seperate supply, to keep the positive rails on a true redundant base, without any other mods.

Re: 300w + 300w = 600w?
 

Im going to have a few friends look at the part with the negative voltages. While they scream at me for trying to make them do that, im going to try and learn enough electronics to figure out what parts needs to be replaced, and by what. If im going to do this, im going to leave the psu's unmodifyed, for warrenty reasons, and it aint happening, unless i can get all the psu's to be hotswapable. I'll go for all 3 running all the time, untill one fails. That should both lower stress, and heat, while if one fails, i just have to keep within the limits of what 2 psu's can actually provide. A concern i have is how to connect the psu to the distributor board, since you dont want to overload any wire's. I might also go for joining multiple 12v rails if i end up using sutch psu's. Dunno if its a good idea or not. its good as it makes sure no rail gets overloaded and pulls the plug, but i understand its bad due to fan noise, and similar causing fluctations in the current. Joining the rails, and splitting it again with a capacitor for each type of device?

Jupp, im confident about this: im in way over my head.
B!

Re: 300w + 300w = 600w?
 

I hear you about the wiring.

I'll be building a frame so that the connectors I got can mate together.

I opted to cut off all the molex connectors, and join all of the same lines together to a single soldered point on a PCB (one for each PSU), which each contain Yo-Duh_87's circuit. The circuit outputs will be joined on a 6 awg bare copper line (using two lines for the ground).

The AC lines will be shared in a similar way.

The power_good signals in my setup will be routed to AND and OR gates, mounted on a seperate PCB, to generate a single power_good signal, where at least two PSUs are running.

Re: 300w + 300w = 600w?
 

Howdy, I'm another late comer to this thread. Basically I'm doing a similar build, trying to make a redundant PSU for a fileserver. When I went seraching, I found Yo_DUH's article and later this thread. First off, I'm amazed at how long this has been going and that you'd been able to stick with it. Second, I have a couple of questions:

1) What are you doing about the negative voltages? ATX connectors have a -5 and -12 rail? I know these are low current so you could just use the lines from one PSU, but that kills the redundancy.

2) Why do you need the complicated FET drivers and the headache of SMD? I know many ways you can drive high current 12V from just a TLL (3V) source. There are nicer examples in Horowitz and Hill (page 167 C in the 2nd revision is the best I found, it has a circuit to use a filter capacitor with short circuit protection which would remove any ripple, I'll scan it if you want) but here are some (simpler, though crudly drawn) ones I could find online: http://www.tim.woodburn.btinternet.co.uk/ttl2fet.htm

3) How exactly does the reverse protection work? I'm probably just not seeing something. Power FETs generally have very high off resistance and low bidirectional on resistance. Meaning this offers no reverse protection when on, and a single FET would offer roughly the same protection when off. The regular diode in the circuit also doesn't seem to offer any reverse protection as its role seems to just stiffen the ground for, and with the resistor protect, the zener so as to make it work with the driver.

Finally, the only surefire way I can see to have reverse protection is to use an array of diodes, then using the pots to adjust the voltage rails to compensate. Using extra diodes so the current per diode was well over spec would prevent a forward fail (and probably lower the amount you would have to boost them), and since diodes don't have much resistance, very little heat would be dissipated. Though the ones with 0.3V drops are generally in the 5-10A range, so you'd need an array of three or four diodes per FET to be safe. This diode: http://www.st.com/stonline/books/pdf/docs/3619.pdf is available from digikey for about $1.00 and would probably do the job. It has around 0.3V forward drop at 10A and has 25V reverse protection with less than 0.01mA reverse current. Just a suggestion.

Re: 300w + 300w = 600w?
 

Just as I was about to get back on this project, I'm finding out that I have to move, again...


I opted to have the first PSU provide the negatives. -12 isn't needed, but I believe that -5 is still in use.

The MOSFET should be reversed. That'll all I remember for now, sorry.

Maybe in another two months I'll be able to finish this darn thing... In the mean time, I bought a cheap 550W psu (which has a bad 3.3v line :( ).

Re: 300w + 300w = 600w?
 

I wonder if this is still alive.

Re: 300w + 300w = 600w?
 

Actually, yes, but I'm on the move, once more.

Moving to a bigger place, where I'll have a dedicated room. I'm having the carpet ripped out and replaced with tiles, Tuesday.

To help in testing, I picked up a 500 Watt 0.1 ohm resistor; I should be able to really put this thing to the test.