DIY redundant psu...

[FLECOM]

is it realistic to make your own redundant PSU?

i was thinking you could with like loads of diodes, but, finding diodes rated at some of the rail amps is pretty hard/expensive... anyone have any links to perhaps somewhere this has been done?

i was looking at a compaq PSU backplane and they just hooked all the rails to each other... but that seems like a pretty bad idea becuase if one of the PSU's has a "less than graceful" death it will take out the other psu or the machine anyway

 

[gee]

Redundant power supplies seldom ever use diodes on their outputs. When you're pushing 30-40 continuous amps through a shottky diode with a 0.5 volt Vf, then you're dissipating 15-20 watts in that diode, which requires a big heatsink. Instead they use MOSFETs combined with a special controller in order to perform the switching with much lower loss.

A quick Google search revealed the MAX8535/8536 controllers from Maxim. You might be able to through a bunch of these on a circuit board with some big MOSFETs and some other parts, to create a box for making any two ATX power supplies redundant. The uMAX packages are a pain in the ass; check out ON Semi and TI to see if they have any similar parts.

Chances are, your Compaq supplies have circuitry like this on their outputs.

When you hook these things up, make sure that the two supplies have *exactly* the same output voltage - otherwise one supply might take the entire load and the other supply might go unstable. Normally, redundant power supplies have current sharing lines going between each other to ensure that this doesn't happen.

 

[Cardboard Hammer]

Perhaps solid state relays could work. Put a relay on each output line of both PSUs. Splice the output side of each pair of relays together. Rig the relays of the primary side to stay closed while the primary functions and to open if the primary fails. Use voltage from the primary to bias against the secondary trying to close its relays such that the relays stay open until the primary fails, at which point the relays close and the secondary PSU takes over. There might be issues during the cutover, quick though it should be... perhaps it'd be possible to buffer things with capacitors and / or coils after the merge.

 

[Ice Czar]

is it realistic to make your own redundant PSU?

not unless your an electronics wiz
I seriously investigated that

synch 2 PSU each as a seperate bus

1. Hardwired
2. With a relay

or employing a shared bus Like a redundant PSU

1. with resistors
2. with mosfets
or how its really done
3. with diodes (Single Source Fault Tolerant Power Systems) @ Power Electronics Technology

I own several redundant PSUs and ran a Hardwired pair for over 2 years
the real way which is hotswappable is not easy to accomplish and given the sensitivity of modern mobo power schemes its the only one Id trust

Redundant power supplies seldom ever use diodes on their outputs. .

your sure? thats the scheme in that last link, its slightly dated, but...

The cost of configuring the above 1+1 system is more than double that of a single power supply of 1200W. Overall cost of this system includes the second module plus the isolation or “ORing” diodes, and a mechanical enclosure that houses the two modules. The Photo 1, above, shows a redundant system of 1200W made of three 600W modules. Note the added complexity compared with a system containing just one power supply module.

Inclusion of the ORing diodes in series with the output is critical to the N+1 configuration. They're not optional if the objective of the paralleling is redundancy. These diodes add cost, voltage drop, and heat — necessary penalties. These diodes guarantee no matter what type failure occurs in one of the power modules, it won't drag the dc bus down. Such a failure can be any component open or shorted (on primary or secondary side), including shorted output capacitors in the module. No truly redundant power system can be without ORing diodes.

Schottky diodes are advisable for the ORing function because they have a lower voltage drop compared with conventional power diodes. However, this drop may be too high for a 5V or 3.3V output with current over 10A. In that case, a FET with low on-resistance should be used instead. A 5mΩ FET drops only 250mV for a 100A output in contrast with 750mV for a Schottky diode. The power saving is 50W, 10% of the output power, and a potential problem in high-density packages.

Use of an ORing FET can be tricky, due to its ability to conduct in both directions — especially if the module employs synchronous rectification. Here, the functioning modules in the N+1 system can pump current into the failed power module as long as the ORing FET permits such reverse current flow. The synchronous rectifier in this mode acts as a booster to pump energy from the common output dc bus to the primary side of the failed module. As a result, the extra current drain on the remaining redundant system modules may cause sagging or overheating.

You must ensure that the ORing FET does not get any gate voltage if the module fails, causing it to conduct in reverse. To avoid this problem, use an arrangement that produces the gate voltage for the ORing FET from a dedicated secondary of the power transformer. If the module fails, shuts down, or is willfully inhibited, then the gate can't receive any drive and the FET will be turned off.

there is alot more, I gathered that multiple schemes where employed, depending on the rail

 

[gee]

Perhaps solid state relays could work. Put a relay on each output line of both PSUs. Splice the output side of each pair of relays together. Rig the relays of the primary side to stay closed while the primary functions and to open if the primary fails. Use voltage from the primary to bias against the secondary trying to close its relays such that the relays stay open until the primary fails, at which point the relays close and the secondary PSU takes over. There might be issues during the cutover, quick though it should be... perhaps it'd be possible to buffer things with capacitors and / or coils after the merge.

Soild state relays have too much of a voltage drop. It depends on what they're designed to switch, but generally there's a back-to-back pair of SCR's inside and the voltage drop is way beyond that of a diode.

Real relays won't switch fast enough to provide effective isolation, compared to an ORing diode or mosfet arrangement.

your sure? thats the scheme in that last link, its slightly dated, but...

I'm certain about this; MOSFETs are used in most applications today.

ORing using diodes is actually *more* complicated than using fets. Since the voltage across a diode varies with current, redundant power supplies have to take two voltage samples back - one from after the diode to ensure that the redundant bus voltage is correct, and one from before the diode to keep the power supply stable.

With a good fet based OR controller, the power supplies don't have to regulate the redundant bus voltage, since the voltage drop across the fetss is almost insignificant. You can even buy ultra-low-Rds(on) MOSFETs specificially for ORing applications, which have a huge gate capacitance making them useless for switching applications, but an obscenely low on-resistance.

Also, ORing diodes operating at high voltages will put off a *lot* of heat, requiring either an obscenely large heatsink (which affects mechanical design) or a smaller fan-cooled heatsink (fan = low MTBF).

 

[Ice Czar]

damn there goes the validity of that link

you killed it, any replacements?
Id really like something describing the current tech in N+1
or a clue where I might go search

 

[gee]

damn there goes the validity of that link

you killed it, any replacements?
Id really like something describing the current tech in N+1
or a clue where I might go search

Read a few datasheets... look for the Maxim chip I mentioned earlier, and here's a similar Linear part - http://www.linear.com/pdf/4351f.pdf

You can probably find a few appnotes for these chips, too...

 

[FLECOM]

well i want to get two of the fortron 550 watt psu's and run them in a redundant/load sharing config... i dont care about hot swap as it isnt realistic to make a hot-swap cage...

the only question i have are, are there any fets that could realisticly handle those current loads?

also are components like that available in non-surface mount? like just regular DIPs?

not unless your an electronics wiz

you doubting me there buddy?

just becuase i havent done a lot of work with high current powersupplies dosent mean i dont know what end of the soldering iron to hold

 

[Ice Czar]

you doubting me there buddy?

just becuase i havent done a lot of work with high current powersupplies dosent mean i dont know what end of the soldering iron to hold

Not at all
as a matter of fact, if you manage to pull this off, your going to be the hit of the town
I pestered a number of electronics wizard to attempt this before I finally gave up and bought a bunch of N+1 and 1+1

and considering some of the membership resources you might also be able to get to assist, like gee, I dont doubt youll be able to do it

and once you do. I'll be making one too
like the sound of an $800 supply for less than $250

 

[FLECOM]

buy me my other 550 and i can get started

not that ive gotten the first one yet... sigh... damn visa wanting their money

lol

o well, guess its time to start downloading loads of component PDF's yay

 

[FLECOM]

im looking at the sheets from the maxim parts, and they look almost too good to be true, the circuits from the evaluation kit are painfully simple...

only thing all these parts are surface mount... and i have a complete inability to work with surface mount (i jitter... a lot, dont ask )

the maxim chips ex MAX8536, you put that circuit on the output of each PSU and then tie the outputs of all of the circuits together? (one per psu per rail essentially?)

also the mosfets that the dev kit uses are Fairchild FDB7045L that are rated for 30v at 75A !!!

this is actually really really easy... minus the whole surface mount part... you could probably shove all this inside the psu itself... hrmm... ideas...

how much heat do mosfets dissipate?

 

[gee]

How to figure out static FET dissipation...

1. Find out what voltage the chip is driving the FET with.
2. Then look up the datasheets for the FET you wish to use; there's almost always a Rds(on)-versus-temperature curve in there somewhere. Get Rds(on).
3. Multiply Rds(on) by I^2 to get the power dissipation in watts. Figure out how well you're going to be heatsinking these (calculate a degrees-C-per-watt rise) and estimate the temperature.
4. Look at the Rds(on) curve again, calculate I^2*R again, and recalculate the temperature rise. Keep iterating with this process until you've got a stable number.

Chances are you won't find a DIP controller, but you might be able to find a SOIC chip from TI, National, ON Semi, Fairchild, ST, etc which are far easier to solder than the uMAX chips from Maxim. uMAX are downright painful to work with - they're designed for machine placement and not human placement.

Finding a big leaded MOSFET won't be hard, go through the parametric tables on irf.com and see what you find.

Good luck with things!

im looking at the sheets from the maxim parts, and they look almost too good to be true, the circuits from the evaluation kit are painfully simple...

only thing all these parts are surface mount... and i have a complete inability to work with surface mount (i jitter... a lot, dont ask )

the maxim chips ex MAX8536, you put that circuit on the output of each PSU and then tie the outputs of all of the circuits together? (one per psu per rail essentially?)

also the mosfets that the dev kit uses are Fairchild FDB7045L that are rated for 30v at 75A !!!

this is actually really really easy... minus the whole surface mount part... you could probably shove all this inside the psu itself... hrmm... ideas...

how much heat do mosfets dissipate?

 

[FLECOM]

How to figure out static FET dissipation...

1. Find out what voltage the chip is driving the FET with.
2. Then look up the datasheets for the FET you wish to use; there's almost always a Rds(on)-versus-temperature curve in there somewhere. Get Rds(on).
3. Multiply Rds(on) by I^2 to get the power dissipation in watts. Figure out how well you're going to be heatsinking these (calculate a degrees-C-per-watt rise) and estimate the temperature.
4. Look at the Rds(on) curve again, calculate I^2*R again, and recalculate the temperature rise. Keep iterating with this process until you've got a stable number.

gee [Level 10 PSU God] has hit FLECOM [Level 5 Moderator] for 50 points dmg... roll again? _

so im guessing a giant heatsink from a blown psu with a fan should do?

i honestly have no fcking clue what you said... you lost me at step 2 lol

 

[Ice Czar]

gee [Level 10 PSU God] has hit FLECOM [Level 5 Moderator] for 50 points dmg... roll again? _

hmmmm...now thats some fine sig material, but is there room?

 

[gee]

gee [Level 10 PSU God] has hit FLECOM [Level 5 Moderator] for 50 points dmg... roll again? _

so im guessing a giant heatsink from a blown psu with a fan should do?

i honestly have no fcking clue what you said... you lost me at step 2 lol

example:

http://www.irf.com/product-info/datasheets/data/irl3502.pdf

This is an ideal ORing diode MOSFET; it has an Rds(on) at 25C of 0.007 ohms with 4.5 volts on its gate, and it has a max junction temperature of 150C. If you just stand the FET in the air with no heatsink whatsoever, it has a Rtja of 62 degrees C per watt (bottom of first page). Suppose you put a continuous 40A through the part, I^2 is 1600, and 1600 * 0.007 = 11 watts... and 11 watts * 62 degrees C will bring you way above 150C. So obviously you'll need some heatsinking... Also if the MOSFET gets hotter, its Rds(on) will go up.. this will make its power dissipation go up, and it'll get even hotter than before!

Easiest thing to do is... suppose you want the MOSFET to run at a 100C junction temperature at 30 degrees ambient and 40 amps. Rds(on) from figure 4 is is 1.3 * 0.007, which is 15 watts dissipation. With Rtjc + Rtcs = 1.39 degrees C / W, your heatsink needs to have a thermal resistance of (70/15 - 1.39) 3.28 degrees C per watt.

As for how to pick heatsinks, that's not my job.

 

[FLECOM]

so like an alpha 8045 or something?

if these things are dissipating like 11~15 watts then thats no big deal... i think MOSFET's i think power amps... i think heatsinks the size of my head lol

so i think some PSU heatsinks should do, since they are going to have some active cooling also...

anyway i can dissipate heat, my bigger issue is wtf am i going to do with these stupid surface mount chips

 

[FLECOM]

hmmmm...now thats some fine sig material, but is there room?

you got space

dude 4gb ram? why?!

 

[nst6563]

A quick Google search revealed the MAX8535/8536 controllers from Maxim. .

Speaking of Maxim...anyone get this month's issue yet?? Another golden Anna issue

that's almost enough to make me a tennis fan......




NOT!


sorry...couldn't resist...had to throw in some more useless banter amongst the arguing over PSU's.

 

[Vertigo Acid]

dude 4gb ram? why?!

So he can make a RAM disk of all his pr0n for super speedy access!!!

 

[Ice Czar]

dude 4gb ram? why?!

you just noticed that?

Animation Rendering, not that Ive bought the Rendering Engine neccessary to use it yet
costs more than the computer

still using freeware and trials, few programs are coded to employ more than 2GB per thread and even fewer are long coded for 64bit, but by the time I have the $$$ for the frontend and backend, they maybe

in the meantime I can use a full 2GB of memory per a thread if necessary (photoshop) so memory bandwidth isnt a constraint, real fun to play with that NASA photo everyone used to download as a browser test, unlimited history, shift color profiles, ect, adds up fast

Ive got it burned to CD and load it to the HDD nearly as soon as the computer is up and running
use it as an informal memory and swapfile\scratchdisk test on all of them, the workstation doesnt even notice
well after its been loaded and before I do too many things with it

http://visibleearth.nasa.gov/

land_shallow_topo_east.tif

hmmm no longer see it, and cant log on to the ftp

 

[Harkamus]

is it realistic to make your own redundant PSU?

Funny you should post this. My bro just made his own redundant PSU a few days ago out of two Antec 430s.

 

[Ice Czar]

shared bus? or just synchd?

 

[Harkamus]

shared bus? or just synchd?

I'll have to ask him when he gets back.