300w + 300w = 600w?

[bigben2k]

Good tip, thanks punish3r, I'll check it out.

I went over the Digi-Key options, and I'm not happy with a single one of them.:shrug:

Three PSUs will be mounted in the 5 1/4 bay areas, as soon as I figure out how to trim 1/8" from the sides...

I'll also be wiring in a lit AC switch to each PSU


along with status LEDs for "Stand-by" and "Power_ OK".

I'm also wiring in a pump relay, with a power outlet, not to cut the pump cord.

[Yo-DUH_87]

I was going to ask if you were going to wire your pump into this badboy, I guess you answered that for me!

Those are nice switches, I have one for my next project just waiting...

Have you suceeded in trimming the PSUs down so they will fit in the dead CD-ROM cases?

[bigben2k]

Originally posted by Yo-DUH_87
Have you suceeded in trimming the PSUs down so they will fit in the dead CD-ROM cases?

There is NO WAY that these PSUs can be trimmed down, without some significant work!

The main blocker is that there are 2 PCB that are standing up. One of them is on the side of the PSU, and runs the whole height. The connector is a double row of 0.100" pins, and there must be 20 pins, at least.

The second PCB stands up right in the middle of the main board, near the back of the PSU. It's not as tall, but is also connected with another double row of pins, also of the 90 degree variety.

Other than that, the heatsinks occupy a huge amount of space, but they'll be replaced with a copper plate, with a copper tube soldered on top of it. There's also a transformer that's pretty tall, and a capacitor that's even taller.:shrug:

[Yo-DUH_87]

So, how are you going to cram all these plus any optical drives you may want to use into your case?

[bigben2k]

Unless I start doing some serious case cutting, I'll only have room for the one CD-RW, which'll have to act as a reader.

I've had my eye on this one for a while: Lite-On (to match the PSU ) LTR-52246.

[Puzzdre]

bb2k, can you 'stack' three of those psu's vertically (I mean the long side up, and all three in paralell) like this:

|_||_||_|

think that'll use the space of three 5.25 bays and wouldn't be much longer than average drive that goes in the bay.

It could be a little tight for all the wires, especially power cords, but I think it's manageable?

[bigben2k]

thanks, but it won't help.

The PSUs are 6" wide (a smidge more than a 5 1/4 bay), 5 1/2" deep, and 3 3/8" high.

They *almost* fit horizontally.

I plan to add a handle on each one: I found this black anodized Alu handle at Mouser.com that should fit perfectly (yeah, right!).

I also have a case painting job to do as well as a window mod . I also have to mod the front cover, to allow air to flow out (!) as well as painting that plastic piece . Then I gotta fit the 4 UV CCFLs, oh and make sure I etch something cool in the window

[Yo-DUH_87]

Originally posted by bigben2k
...I also have a case painting job to do as well as a window mod . I also have to mod the front cover, to allow air to flow out (!) as well as painting that plastic piece . Then I gotta fit the 4 UV CCFLs, oh and make sure I etch something cool in the window

You don't HAVE to make it cosmeticaly pleasing, but it is always nice to have something to show off to "mere mortals (wife/gf/parents)" for your spent $$$

Could you mount the PSUs hot-swap somewhere else in the case, such as above where the normal PSU would go? There is a lot of space...

[bigben2k]

There *used to be* a lot of space It'll be occupied by the heatercore (~6" by ~8"), and the blower.

The airtrap will be sandwiched between the blower, and the PSU array's power bus.

The electrical wiring throughout all the case will be very neat, and very short, since everything will be hooked up to the power bus.

[Yo-DUH_87]

Aah, I thought you may be using that space for watercooling, but had to try

[bigben2k]

The switches arrived yesterday, along with the rest of the supplies. Purdy.

I found a handle:

0.75" high. 0.187" diameter steel stock. Shiny dark gunmetal enamel finish. Threaded for 4-40 screws on 3" centers.
CAT# HDL-13 Price: $1.00 each

At the same place: www.allelectronics.com (tip, they have a special right now, where if you place an order of $25 or more, shipping is free: use promo code SHIPFREE).

And to top it all off, I think I found the connectors:

Molex-type multi-pin connector with contacts on 0.198" grid spacing. Contacts are crimped onto wire and inserted into housings. Polarized male and female white nylon housings interlock with detent lock. Can be snap-mounted into chassis cutouts. Crimp pins accept 24-16 AWG wire. Includes 2 plugs and 2 receptacles, 8 male and 8 female crimp terminals.
CAT# CON-33 Price: $1.50 each

Molex-type multi-pin connector with contacts on 0.198" grid spacing. Contacts are crimped onto wire and inserted into housings. Polarized male and female white nylon housings interlock with detent lock. Can be snap-mounted into chassis cutouts. Crimp pins accept 24-16 AWG wire. Includes 2 plugs and 2 receptacles, 22 male and 22 female crimp terminals.
CAT# CON-39 Price: $3.50 each

I'll add a grill, for the front door, to allow airflow.
detail here


[edit: if these contacts are the same as what I found for Molex/GC/Waldom in my Digikey catalog, the contacts are rated for 12 amps each. Exactly what I need.

[pHaestus]

If you have the width to spare, you can desolder the upright PCBs and bend the headers at 90 degree and then resolder. I have done such before

Might also want to consider mounting one of the psus upside down so you cool two with a single pipe/plate setup.

I would suspect you could "make it go" with a bit of effort

[bigben2k]

Originally posted by pHaestus
If you have the width to spare, you can desolder the upright PCBs and bend the headers at 90 degree and then resolder. I have done such before

Might also want to consider mounting one of the psus upside down so you cool two with a single pipe/plate setup.

I would suspect you could "make it go" with a bit of effort

Unfortunately, I've reached the maximum width of the 5 1/4 bays, and making it larger would involve some serious cutting of the case, and I would only gain maybe 2 inches.

I think it'll work very well the way I have it planned out right now. I just need to widen the 5 1/4 bay area a little bit. Once I get the connectors, and put in the backing plate for the back connectors, the PSUs should slide in and out very easily.

I see what you mean about the upside down PSUs, but since I'm aiming to make it all removeable, I'm going to make all the PSUs the same.

I might have to mod the front cover, for the handles to fit.

The next step, once I've got this array up and running, is to figure out how to water cool the whole lot, and how I'm going to have quick disconnects, that hopefully don't spill any. I also have to keep in mind where the tubes are going to come out, in relation to the power connectors.

Other than that, I might order replacement fans for the PSUs:

[bigben2k]

Here's a pic of the case.

The 5 1/4 bay area has walls on either side, but the round holes in it protrude inside the bay area, by at least 1/8", from each side.

I have to somehow trim these enough so that I can slide the PSUs in there.

There are also drive rail guides, at the front of the case, but I can just bend those back a bit, and it should work.

I'm actually thinking about using the drive rails, for a nice fit, but I suspect that each height would require a different rail arrangement, so I just might trim off the original rail guides, and put it my own.

[bigben2k]

Here's a rough layout:

If everything works out well, I might even throw in the flowmeter.

[bigben2k]

Weekend progress report:

While I was messing with the case, trying to figure out the best way to fit the PSUs, I found out that turning them 90 deg, they actually fit. (Didn't someone suggest that?). So the fan blows into the side of the case.

Well...

I completely stripped 3 dead CD-ROM drives, and after a bit more twiddling, I am opting to strip the PSUs completely of their original PSU cases, and mounting them directly on the remains of the CD-ROM drive case. So I have to order more stand-offs

I'll also try to route some of the wiring inside the case support beams. It'll be very clean!

I've positioned the pump relay, and picked up 4-40 screws, nuts, and a tap, for a nice clean fit.


PSU#4 has been chopped up, and will be my prototype, for the wiring fit. I'm pondering wether I should use the remaining wiring to connect the PSUs to the Yo_Duh circuit, or if I should use my 6 gauge copper "wire" instead. There are a lot of wires!
+12 : 5 wires
+5 : 12 wires (including a special one, see below)
+3.3: 5 wires (also including a special one)
GND: 18 wires (!)

The +5 and +3.3 lines include a seperate wire, with the same color, that was wired directly into the mobo header, on the same pin as the regular voltage lines. It looks like they were some kind of additional filtering or voltage stabilizer: they're linked to a cap and resistor.

I guess that this additional stabilizer line will be fed to the Yo_Duh circuit, since I can't run it to the mobo anymore.:shrug:


PSU#3 will be chopped up for a water cool fit.

PSU#1 is my test unit, PSU#2 is the spare.


I also stripped the 3 CD-ROM of all salvageable parts including some small actuator/motors I could re-use for the front cover, maybe.

Most of my fingers survived without a scratch

[bigben2k]

I've kinda split some info between two threads, so here is where I've posted a few notes.

Here's Yo_Duh's circuit.

In my circuit, I've replaced the gate driver with an LTC1163, and the MOSFET with an IRF1404.

The driver will be powered by the "PWR_OK" line, so that the outputs become active once the PSU's outputs are stable.

There will be three MOSFETs per PSU, one for each of the +12, +5 and +3.3 power lines.

I'm also adding a logic circuit, to route a "PWR_OK" signal to the mobo, only when at least 2 PSUs are ON.

Now...

As Since87 pointed out, there may be a flaw in the circuit, where if one of the PSUs is turned off, it may be fed by the other supplies. Otherwise, as long as the PSU outputs are pretty close to each other, everything should be fine.

I proposed reversing the Source and Drain pins of the MOSFET, as a cure.

Also, on the zener diode (15v zener in the article), it would allow a voltage to reach the MOSFET gate, but too low to activate it. So I've modded the circuit to have the zener wired to ground, instead of the Source.

Now all I have to resolve, is how to make sure that the driver is going to output enough voltage to drive the MOSFET for the +12 line.

I'll also be reviewing some added circuit protection, from the IRF website. I have to assume that the load (PC) is capacitive/resistive, and not inductive.


QUESTION:
Incidentally, I have a stupid question: since the zener has to be at least 16v, to activate the +12 line, and no more than 23.3 to activate the +3.3 line, can I use a combination of two zener diodes in series, one rated at 16v and another rated at 3.9 (for example), to achieve what I need?

[bigben2k]

Almost forgot...

I also outlined a number of steps to follow, to test the array as it's being built.

#1: run 1 PSU with one circuit (with a test load)
-measure driver output voltages
-verify output voltages

#2: run 1 PSU with 2 circuits (1 ON, 1 OFF)
-measure MOSFET gate voltages

#3: run 2 PSU with 2 circuits (all ON)
-measure PSU output voltages
-measure power bus voltages

#4: test "PWR_OK" circuit

#5: route -5, -12, and +5standby from PSU#1 to mobo header, as well as "PWR_ON" from all three PSUs.

#6: wire in lights (and switches?)

#7: wire in peripheral connectors, and pump relay.

Then all I have to do, is build up the CD tray/case for the connectors, finish attaching the PSUs to them, build a frame within the case for the receiving connectors, wire it all in, and route everything neatly.

[bigben2k]

Information update:

The zeners in series work. Since I already got 16v zeners, I'll add a 3.3v zener for the +12 lines, limiting the driver output to 19.3v, and giving me a max Vgs of 7.3v .

From the datasheet of the 1163 driver, it's suggested to add a 150 ohm resistor to the supply pin, if negative transients are expected. 300 ohm is recommended for the 1154 driver.

From IRF:
Gate Drive Characteristics and Requirements for HEXFET® power MOSFETs (PDF, 21 pages)

There are a number of ways suggested to drive the gate of a MOSFET.

IRL have a min Vgs of 1 to 2v, instead of the IRF's 2 to 4, to turn on the MOSFET.

The Do's and Don'ts of Using MOS-Gated Transistors (PDF, 7 pages)

From that, I can see that my suggestion to reverse the Source and Drain, would actually allow the PSU voltage to get to the power bus, with a certain voltage drop, regardless of the state at the gate. Back to the sketchboard.

I can also see the detail of the purpose of the zener: the transient to the gate comes from voltage drops at the source/drain.


Note to self, for the next time I do this: consider IRF2804 or IRFP2907, or a good IRL.

[Since87]

Originally posted by bigben2k

The Do's and Don'ts of Using MOS-Gated Transistors (PDF, 7 pages)

From that, I can see that my suggestion to reverse the Source and Drain, would actually allow the PSU voltage to get to the power bus, with a certain voltage drop, regardless of the state at the gate.

How so?

I suspect you are misinterpreting something.

If the source is tied to a PSU, the drain to the power bus, and the gate to ground, current into the PSU output will be less than 20uA. (Idss spec)

Ben, your design has lots of substantial flaws compared to Yo_Duh's, but turning the MOSFET around, while unorthodox, is not one of them.

Here's a hint, read ALL of this line from the MOSFET datasheet:

VGS(th) Gate Threshold Voltage 2.0 ––– 4.0 V VDS = 10V, ID = 250µA

Look at Figure 1 of the 1404 datasheet. This data is for using the MOSFET in an orthodox manner. Using it 'backwards' is likely to be 'worse'.

[bigben2k]

Thanks for the vote of confidence

As far as I can see, the original problem was that the current could flow back into a PSU that was turned off, because current could flow through the zener backwards, with a minimal voltage drop (up to 1.3).

So by switching the drain and source, all I'm doing is switching the problem around: the PSU can now feed the power bus, with minimal voltage drop (up to 1.3) through the same diode, the same way.

plus... in this backwards configuration, when the MOSFET is actually activated, the "minimal" voltage drop actually goes away.

Which part of the above is wrong? What am I misunderstanding here?


Otherwise, I'm only aware that I may not get enough voltage to drive the MOSFET for the +12 line, on each PSU, but I've already mentionned that. Since Vgs needs to be 2.0 to 4.0, given a 12v supply, the gate voltage needs to be at 14 to 16 volts, and preferably higher, where the driver output appears to be clamped at 14. Yo_Duh and I discussed that, and he assures me that the output will be sufficient, based on the maximum rated voltage increase of the output over the supply, of 13v which, above a supply of 5, would output 18v: plenty to drive the MOSFET of the +12 line.

I already knew that.

As for the current, I have yet to confirm that the driver can actually supply that, yet Yo_Duh tells me that it could.

So what else am I missing?

My solution to the above, which I have yet to verify, is to swap the MOSFET for an IRL1404, instead of the IRF1404, only for the +12 line. The Vgs then drop to a range between 1.0 and 3.0, which is slightly less of a probleem, but still a problem.

Either that, or build my own driver from scratch.


I'm also considering leaving the circuit with the MOSFETs in the right (orthodox) placement, and forcing all the drivers to turn on all at the same time, once and only once all three PSUs are "ready", which would effectively mean hardwiring the PSUs together.

I really don't want to use the MOSFETs in an odd way, like I've suggested.:shrug:

[Since87]

Originally posted by bigben2k
Thanks for the vote of confidence

And what confidence do you expect? You started this thread nearly six months ago and still don't have solutions to fundamental issues.

Originally posted by bigben2k

So by switching the drain and source, all I'm doing is switching the problem around: the PSU can now feed the power bus, with minimal voltage drop (up to 1.3) through the same diode, the same way.

I'm not sure what the problem is. I thought the goal was for the PSU to feed the power bus when it is on and working correctly.

Originally posted by bigben2k
Otherwise, I'm only aware that I may not get enough voltage to drive the MOSFET for the +12 line, on each PSU, but I've already mentionned that. Since Vgs needs to be 2.0 to 4.0, given a 12v supply, the gate voltage needs to be at 14 to 16 volts, and preferably higher, where the driver output appears to be clamped at 14. Yo_Duh and I discussed that, and he assures me that the output will be sufficient, based on the maximum rated voltage increase of the output over the supply, of 13v which, above a supply of 5, would output 18v: plenty to drive the MOSFET of the +12 line.

I already knew that.

Ben, you still don't know how to interpret the Vgs(th) spec. That spec is telling you that the part is only guaranteed to pass 250 microamps when the gate voltage 4V above the source voltage AND the drain voltage 10V above the source voltage.

The spec to be concerned with for the LTC1163 is the MINIMUM Vgate-Vs. That value is 5V. You can't count on the LTC1163 getting the gate more than 10V above ground. Even if you count on getting 'typical' performance, and get 14V out of the the LTC1163, that is nowhere near enough voltage to guarantee that a useful amount of current will flow through the MOSFET.

Originally posted by bigben2k
So what else am I missing?

A sufficient level of understanding of the issues involved in designing this.

[Yo-DUH_87]

First of all, I want to apolagize for being away. I had been hoping that the article would get published a while before I left, so that I could answer questions. I did have a productive trip, but I am totaly zoned at the moment (due to the lack of sleep that usualy accompanys traveling), so I may not make total sense

The diode in the mosfet is not a zener, it is a body diode. Yes, it theroticaly causes leakage in reverse, but as my circuit was intended for both PSUs to be on at the same time, not what ben is doing, while assisting ben I totaly overlooked the body diode. Thanks for pointing that out.

I may fire up the prototype circuit that I built to get some real-world results. If I remember correctly, durring testing and actual usage, I could turn one PSU off, and leave the other on, no problem. I don't recall reading any back-voltage, but it was awhile ago (about 2 months now), so we'll see.

When I was assisting ben with the design, I did look at the mosfet drivers specs. It is rated for 250 microamps, the mosfet power drain is around 9 nanoamps. That should allow sufficent head-room. Those are prety powerful drivers there. The drivers I used I believe had a current of around 10 microamps (again, check me on that), and put out about a volt over typical.

Ben, I asked my father about a solution to your problem. You had something with your reversed mosfet idea, and he suggested a solution. I'm going to do the testing first, though.

Since87, are you and BillA related in any way

[Since87]

Originally posted by Yo-DUH_87

The diode in the mosfet is not a zener, it is a body diode.

Hey, hey, hey. That's getting extremely picky. Uh, err.... nevermind.

You're right, I was speaking sloppily in calling it a zener. The fact is though, that for practical purposes, it can generally be treated as a zener. That is the reason the symbol for a zener diode is shown within the MOSFET schematic symbol.

Originally posted by Yo-DUH_87
[b]
When I was assisting ben with the design, I did look at the mosfet drivers specs. It is rated for 250 microamps, the mosfet power drain is around 9 nanoamps. That should allow sufficent head-room. Those are prety powerful drivers there. The drivers I used I believe had a current of around 10 microamps (again, check me on that), and put out about a volt over typical.

The output current of the driver is not much of an issue for steady state operation. (It's a big issue at the time of switching.) Assuming the load can tolerate a slowly rising rail at powerup, the main issue is the output voltage.

Designing around the performance of particular parts on hand is fine if you are just building a one-up version of something for yourself. If there is any possibility that you might promote the design to others, it would be a good idea to design according to the worst case specs provided by the manufacturer. The LTC1163 is totally inadequate for designing a circuit for switching 12V. That's not what it is designed for. The fact that a person may get it to work, definitely does not mean another person can expect to get the same results.

Originally posted by Yo-DUH_87

Since87, are you and BillA related in any way

I don't know, but I suspect Bill may also be a member of the brotherhood of Men United in Chastising Otherwise Unchecked Stupidity. Pteww. (Spit when you say that.)

Doesn't really count? Well then if you go back about 100,000 years...

[Yo-DUH_87]

When ben and I were discussing the driver, I was expecting ben to have the extra boost from the zener diode connected to the source of the mosfet. That would have worked out quite nicely (as it did in my circuit). But, alas, ben had to bring up the "small" issue about the zener possibly allowing current to flow through the zener and power on the mosfet when you didn't want it on, so the quick solution was to connect it to ground...

@ben, I'm working on figuring this all out. Hopefuly I'll have some time this afternoon