300w + 300w = 600w?

[murray13]

bb2k: Two things: 1. You can and you can't just hook two power supplies up together without running a BIG risk, without knowing how computer switching power supplies work in detail. If anyone has a redundant P/S to look at there is more inside than just two supplies. There is circuitry to load balance between the two supplies. The supply rails ARE just hooked together. BUT the voltage sense lines go thru a circuit to ensure that each supply is putting out the correct voltage. If you don't have that circuit then one supply can actually become a LOAD on the first supply, and that is not good. I could go into more detail if you would like, but I think the best idea other than finding a very lg P/S cheap is to split your system up (one P/S for MB, one for HD, CD's etc.)
2. I have two redundant P/S's I'm not using. They are HUGE tripple P/S's rated at 660W. They have three slide in supplies with one in reserve. Hot swappable. But they are big, almost the size of a mini-tower. And the downside is they need a new wiring harness.

If you would really like to know some more about why it's not a good idea to actuall hook two supplies together, let me know.

[bigben2k]

Ok, I'll bite. Why isn't it a good idea?

I hear you about the load balancing. That's my main issue with doing this, along with what happens if one of the PSUs go dead (as unlikely as it would be).

As for a load balancing circuit, my issue with that is:

a) it's almost the same as building a PSU from scratch (but hey, if that's what's needed...)

b) it would require higher voltage, in order for the output to be within the range that it needs to be. (it could be tweaked out of the PSUs).

I figure that if I pick the same model PSUs, that I tweak the output to match, as closely as is possible, I should be able to avoid having to put a load balancing circuit. However, it doesn't address the dead psu issue. If I add a diode at the output of each PSU, it should do the trick, but then I'm looking at tweaking up the outputs by 0.7 volts (see B)

As far as mounting goes, I don't have a solution yet. I only want one power cord coming out of the computer, so that could be a bit tricky. What I do know for sure, is that these cheap PSUs are going to get cut up, so that I can convert them to slimlines. I can put together some kind of alu rail-rack.

Aaaargh! I need a PSU diagram!

[FARMER]

Well i have my mobo hooked up on a 300watts, and all the other stuff are on a 250watts AT.
The site is in danish so you won't be able to read any of it well ripped out the back of the tower and made a new one (for dual psu, one 92mm blowhole, and room for the ps/2 wires) welded the mother****er back on and painted it

The 12volts(from the ATX psu) turns on the 230volts for the AT. This is done by a little thing which I don' know the english word for

Here's a pic:
http://www.webbyen.dk/vishjemmeside_...webside=678020

and my computer is NOT messy, just an illusion

[murray13]

bb2k: It's a chip designed specifically for sharing power supplies. It get's put into the regulation circuit for each P/S and controls the output of each at the same time. I looked into adding one of these between a couple of cheap P/S but there is just too much modification to each P/S to take the time to do it.

IMHO personally I would just search the web for the cheapest big P/S I could find and call it good.

btw have you added up the power req. for all the devices in the system to see what you really need? ...unless your going to run a pelt or 6+ really big HD's 450-500W is usually more than enough to be stable.

[bigben2k]

It's not so much a question of what the power requirement is... as it is a question of running the PSU(s) cool and steady.

Do you have any info/link on this "chip designed specifically for sharing power supplies"?

[murray13]

Here is a link to the pdf at Texas Instruments web site:

http://www-s.ti.com/sc/ds/ucc39002.pdf

as it is a question of running the PSU(s) cool and steady.

Well, from my observations over the years, especially with power supplys, one good unit running at 90% will almost allways outlast two cheap units running at 68%. Key word here is GOOD supply.

Also about running cool...it all boils down to the efficiencys of the P/S's. A 450W running at 90% capacity with a efficiency of 87% will put out 52.65W of heat. Two 300W running at 67.5% capacity with a efficiency of 87% will also put out 52.65W of heat. In both cases they supply 405W of power. But remember that is just the theory. In real life the two 'cheap' P/S will probably not be 87% efficient, you would have the additional fans running from the second supply, and you cut the MTBF (mean time between failure) in half by using two instead of one. By having two instead of just one your twice as likely to have a failure. I will admit though, all other things being equal, two P/S running at 2/3rds power should in theory run longer than one running at 90%.


If you decide to integrate that IC between two P/S, I'd love to help. It would be nice to document the process and post it all over the place. Providing a 'poor man's' "server" supply.

Whichever way you end up going I look forward to seeing the results.

[bigben2k]

I'm interested.

Let me read through the 50 odd pages that I printed.

In the mean time, I took a quick look at the 3907 chip, but not sure what its function is yet...

Also, I hear you about having to put one of these in each PSU, but I can't help thinking that there's a better way. Maybe not. Since the objective of this exercise is also going to include stability (as in voltage stability), I think this might be it.

If we're going to do this though, we'll have to cover the redundancy aspect, and that means 3 * 300 Watt PSUs.

Are you still up for it?

[reclone5]

I have been wondering the same thing myself, I was trying to run 2 PSU's in parallel to power a tec, fans and other stuff. The first was a 250W AT and could run the tec fine by itself, but when i added a 300W ATX in parallel, the 300W ATX would overload or something(turns off, and makes that strange crackling noise, like when it gets shorted) Prolly doesn't help you any but that is my experience.

[Can O' Beans]

I think the only safe solution if you HAVE to parallel 2 PSUs is to use diodes on all the lines.

The 0.7v drop is an issue, but hopefully you can tweak the outputs just enough so it isn't a problem.

If I had the two PSUs handy, I'd try the diodes and see what I could get. It shouldn't hurt the PSU in any way I can think of currently. Can try it with a few fans as load and see what voltages you get.

If the lines are high, like the 13v mentioned before, resistors can help this. I had to resistor an old PSU due to a 5.89v output under load.

[Stepback]

Hey Ben,

You might want to stay away from that PowMax power supply I pointed you to earlier in this tread. Not a day after I posted that it took a SH*T on me. The cooling fans locked up and where not cooling the supply as it has thermo controlled fan I didn't notice it till in went in to meltdown. I've been with out a computer for almost a week waiting to get paid so I could buy a new one. I ordered a Antec Turepower 480 from Directron this morning and the wife is going to stop by there and get it on her way home today.

[bigben2k]

Thanks for the tip. If I pick it up, I'll make it a point to swap the fans.

I stopped by the place a couple of weeks ago, but they're closed on Saturdays... (I was in the area).

[Stepback]

Hey Ben

What part of town are you in? I'm up here on the northside in Spring.

[bigben2k]

NW.

[Can O' Beans]

I got the 550w PowMax PSU for my current case mod project. Since thats not done, I put it in my Linux case - 2x366 Celery at 550 (for now..) I'ts been running fine for about 2 weeks so far...

[Stepback]

Ben: Cool, By the way I final got my machine back up.

Can O, Beans: I would replace the fans in as soon as you can. The fans in my died after six month. Directron still sell it but they try to steer people away form them.

[bigben2k]

Not much progress to report.

I e-mailed birman, and the pics to the slimline PSU article are back up. Note: although he writes that it should work for any PSU, I doubt that it would work with an Antec (but hey, if you have an Antec, then you're not reading this!).

I also figured that since I have 6 * 5 1/4 bays in my DA-01 case, that I might as well go ahead and use three of them for the 300W PSUs. They could become removable from the front, if I want to go that way.

In the mean time, I've got to spec out some power connectors (for easily removal) and figure out how to put the bus bar together. I'm also thinking about a front panel switch (to turn off each PSU) so that it can be removed while the system is running. I'm not running anything critical, so it wouldn't be of any use, except for show.

Also, I plan to use arrays of 40mm fans, coming out the back of each PSU, unless I get into watercooling, but that might be a little more than what I want to get into.

A big bonus is the extra space I'll have in the top/back of the case. I could now fit pretty much any heatercore/fan combo.

[GigaFrog]

Definitively, I would split my system in 2 separate (not wired) power supplies. One on the mobo, and one dedicated to electro-mechanical peripherals (hard drives, CD-ROM, fans ...).

The big advantage of this is that the noise coming fron the hard drives spinning, fans, and all other electric motors put a lot of EMI on the power lines.
When I was tuning my PIII system, ( ~ 2 years ago, time for a new system!), at the highest stable speed, it would crash on hard drive access, right when windows starts loading. I put electolytic caps on the HD connector, and it suppressed the problem.

You can even be more cost efficient by getting a real good well regulated PSU for the mobo, and an old cheap AT for the drives. You may have an AT laying around somewhere, don't you?

By wiring 2 PSUs together, you may run into problems. I wouldn't try. You might even get more ripple on the voltage lines because the 2 regulations will interfer with each other.

With a split system, you might get a higher overclock for cheaper and much simpler.

Good luck.

[bigben2k]

Thanks for the tip Gigafrog!

The redundant circuitry appears to be more complex than I originally expected, and I haven't found a simpler way to do n+1, yet.

I agree, the multiple PSU might add to the overall ripple in the output lines. On the other hand, if I was able to reverse the polarity on one of the PSUs (assuming 2 PSUs), I would be able to eliminate the ripple entirely, at least the one coming from the 60 Hz cycle, but since I was headed for 3 PSUs, I was back to square one.

I'll ponder this one a little longer.

[punish3r]

bb2k, I just skimmed these posts so if someone already answered this, then I apologize. The reason you get 13v on crossing two PSU's @ 12v each, is because computer PSU's are switching power supplies. By connecting the outputs of two supplies, you put both inductors in parallel, there by changing their two individual values into one different value, and thereby change the values present in the LCR series time circuit. Not to mention changing the resonance of the incoming AC ripple by bleeding this 'new' value back through the kickback diode, and changing the value of the output capacitors. So, for an arbitrary value, let's say each PS is using the same 100uH coil. You take both of these, and since they're in series: L^t= [(1/L1) + (1/L2)] (I'm not including mutual inductance here on purpose) so if L1 and L2 equal 100uH we get L^t=1/50 uH which changes the output value of the LCR, but just enough to screw up the compensation by the output capacitors. So the capacitors do their best at an attempt to limit the output to 12v, which they can't because the values have changed, which the monitoring circuit doesn't know, so it keeps sending pules of a specific width and power, and it becomes impossible for the monitor to accurately control the voltage because so many variables have changed. Not only that, but the monitor can't even 'see' the new voltage because of its location within the circuit.
Also, if your PSU design included an output voltage limiting comparator (fed off of a reference voltage) the reference voltage would be wrong as well, because of the changed values of the inductors in relation to the input transformer. So anyway. Just thought that might help.

One thing of note on your UCC29002/39002 chip. It's designed for about 10mA. So, suffice to say, it would be severely lacking for this project. (that'll be the day... "10A of cross-regulation from a .5" SMD package!!!")

[bigben2k]

Thanks for the tip, Punisher! I figured something was up, but I couldn't nail it.

As for the regulating circuit, it is designed specifically not to be in the middle of the power output. Very much like a voltage regulator, all it does is set the voltage, and lets a number of transistors actually handle the power.

I still have a lot of work to do on this one; I'm nowhere close to a solution. I'll have to pick up a 300W PSU, just to have something to play with.

For some reason, PSU schematics are hard to come by. Does anyone have any?

[novaflare]

if all your wanting to do is use one for fans cd rom etc why not jsut get a old at 250 or 275 watt psu it will be more than powerfull enough and if your case is big enough you just need to dremel a place for your power cord to go and drill some mounting holes.
And put the switch on the back of the case.
I ran a set up like that for a long time the entire time i used my inwin q2k i was running a second psu that was old style atx with the power cord pass through no less and i used the jumper power cord to power the main psu i ran all my fans and my dvd rom off of it my hds and cpu fan mobo etc off the main atx fan.

Fans i ran on the at psu was 1 120 4 80s and a 92 mm all high cfm fans pluf dvd rom and new q bay amp.

[bigben2k]

That would be too easy

Besides, I'm going for the slimline PSU mod. All PSUs should be removable from the front, for a quick hot-swap (hence n+1).

Does anyone have a PSU schematic?

[bigben2k]

I found this 300W Lite-On PSU. It was destined for a Compaq server, which is kinda neat. But...

Since I'm doing the slimline mod, I'm worried that this "heavy" PSU wouldn't be mod-able.

On special, for a while at 18.99, or 2 for 36.00 (USD). I'd be ordering 4, holding on to one as a spare.

Note how the above reseller rates the PSUs according to weight (!) and how this unit has already been stripped of a couple of extras (volt switch and, I assume a fat transformer).

An interesting theory...

[pHaestus]

That is the same PSU resold by Fortron that did so well in the THG testing. Sparkle uses big heatsinks, and enormous capacitors. I would be surprised if it would fit into a cdrom.

[novaflare]

like ph said alot of heavy duty parts in a heavy psu.
And well heavy duty partys are generaly bigger than light duty parts so more weight.
You know what wont suprise me seeing some junk manu fature like add lead to their junk psus to make them seem like a higher quality psu lol.