Yo-Duh_87's PSU Load Splitting goodies
 

Yo-Duh_87's PSU Load splitting article is up! Many of you remember OnDaEdg's article that covered this before, but Yo-Duh's method is a deffinite departure from Onda's

This is a definite article for those who are looking for cheap power in a power hungry system and don't want to pay the big bux.

Check it out here!

This is useful information indeed.

Umm.

Unfortunately, there's a fundamental flaw.

Here is the datasheet for the IRF1407 MOSFET.

If you look at the schematic symbol shown on the first page of the datasheet, you will notice that it is different than the schematic symbol shown in the article for the IRF1407.

In particular, the datasheet shows a symbol for a zener diode built into the MOSFET. (The black triangle with the 'bent and rotated Z' on top of it.)

The forward direction for this diode is such that it will allow current to conduct from the load point back into a supply.

This invalidates one of the major pros listed for this design:

"no possibility of the supplies attempting to charge each other"

Very nice article.

One thing you might want to look into when building your own supply is buying a prebuilt DC power supply. Someone I know discovered it was cheaper to buy a premade supply with the transformer he wanted inside of it, then ripping it out, than buying the transformer by itself :D.

That would assume that the voltage difference is greater than 1.3v: if your supply is off by half of that, you had a problem to start with.

I'll let Yo_Duh complete the answer here, as I'm planning to switch off one of three supplies.

That 1.3V number is the forward voltage of the zener with 78 Amps through it. The forward voltage will be substantially lower than 1.3V at lower currents.

My impression, was that part of the goal was to prevent current from flowing back into a PSU in the event of PSU failure or one of the PSU's not being turned on, as well as preventing one PSU from charging up another's output at powerup, due to differing startup times.

I still say the statement, "no possibility of the supplies attempting to charge each other", is false, unless some exceptions are listed.

Edit: punctuation

I am going to be on vacation for the next few days (until Wendsday). I don't have time to reply right now, but I will reply as soon as I get back...

FWIW...


I'll be using the MOSFET# IRF1404, and driver#LTC1163.

The driver has a triple output, and will drive the MOSFETs for the +3.3, +5 and +12v lines. Who needs -5 and -12?

I'm also making a logic circuit to route the power_ok signal to the mobo only when at least 2 PSUs are on.

I've still got some tweaking to do. The 1163 may not output enough voltage to drive the MOSFET of the +12 line, I dunno yet.

Also, I modded the original design, so that the zener is wired to ground, instead of Vs, because it would otherwise allow a MOSFET to be turned ON, from another power supply.


edit: the driver will be powered from the "Pwr_OK" line, which turns on once all the PSU outputs are ready.

edit: fix does not work, removed.

edit: fix does not work, removed.

The solution to what?

What do you think this diode is going to do for you?

Does your design process consist of tossing out random ideas and hoping people will correct the stupidest ones for you?

The fix is for preventing the PSUs from charging each other.

The fix is also documented on the IRF website, but it's for another purpose.

If you don't believe it will work, then tell me why. Otherwise, I'll confirm it with my almost complete circuit, and let you know, OK?


Note: the fix only applies to the 1404 MOSFET: I haven't checked to see if it's ok for the 1407, yet.

Oh shoot!

Ignore my above posts: it won't work.

The internal diode is still going to leak.:mad:

Total time till oath breaking 2:09.

Let the ganging up begin.

Eh, I made a mistake: thanks for pointing it out:rolleyes:

Here's a PROPOSAL, that I'm putting up for DISCUSSION:

Switch the Drain and Source.

That will probably work. The specs on the datasheet won't be too relevant, but the IRF1404 is major overkill for the job, so it's not likely to be a problem. (The fact that you aren't switching the MOSFET off and on repeatedly, also helps in making a lot of the specs unimportant.) Only testing will tell.

You need to rethink the use of the 15V zener diode. Its purpose is to protect the gate of the MOSFET from blowing out under transient conditions.

Absolute Max : Vgs = +/-20V

That's what I thought. The max rating of amps is on the +5 line, at 40 amps, but its shared with the 3.3 . Not a worry.


The zener is something that Yo_Duh and I discussed extensively.

As it is, if you turn off one of the supplies, the other PSU will actually drive the MOSFET of the other PSU, right through the zener. (Is it a problem? I dunno :shrug: )

So my solution was to wire the zener to ground, instead of having it wired to the Source.

Since Vgs needs to be at least 2.0 and max 4.0 for the MOSFET to turn on, with a max supply voltage of 12v, I'd need to change that zener to at least a 16v. With the minimum supply voltage of 3.3 the gate voltage cannot be more than 23.3v, otherwise it'll blow the gate. So a zener that's between 16 and 23.3 should work.

But otherwise, you're still right: the zener is designed to protect the MOSFET from transients: I will certainly review how my mod affects that protection.



I'll post more thoughts later.:cool:

Ok, the MOSFET being back-driven isn't an issue for the MOSFET, because Vgs would be ~ -0.7 (negative 0.7), which won't allow anything back into the PSU that's off.

However...

My 1163 gate driver can handle the voltage applied to its output pin, because it's got a 14v zener there, internally. The 1154 however does not have that internal diode. Effect? I still dunno:shrug: