Server power supplies in series!

[auraslip]

EDIT:

I finally got around to writing an article about this on my site https://sites.google.com/site/shelbyelectro/Mods-and-technical/batteries/server-psu-charger
It condenses this thread into something more readable. Enjoy.


I was inspired by LFP to investigate these mythical server power supplies; cheap, powerful, and reliable! Of course it took me awhile to figure out a simple, inexpensive way to do this. boy did I luck out!

First off I found these puppies for $5 each SHIPPED. I now have eight of them. They are rated at 585w and ~60a on the 12v bus. That's a WHOLE lot of power for $40. Basically, I'll be using this thing to charge what ever EV I can build for the next few years.


The idea is to put them in series to get near the voltage I want, and then use a lower voltage voltage and current adjustable power supply to fine tune the charging voltage and limit current. A s-350-12 meanwell would work well here as it's rated at 30a and we can safely limit the current on it. No where near as nice as the variable 50a unit above, but it's also not $800. A genuine meanwell goes for $60 used on ebay. If I put two in parallel, I could get 60a from this charger. And since that'd be 9 12v units (108v) I could theoretically get ~6.5kw out of this setup. And for only $160


Here is a first draft of the box I'll put them in. I'm gonna get some panel meters to get a clue what's going on with the charging process.


Here is a unit disassembled.

 

[auraslip]

Issues:

Safety:

The AC ground will float on all the units except the meanwell. This has it's own safety issues if something should go wrong and supposedly could potentially show AC voltage on the cases of the PSU. So I'll completely isolate them in a box so that only the power switch and charge leads show.

The PSUs aren't isolated like the meanwell. This means I'll need to keep their metal housings isolated from each other. I'll do this with wood.

I'll need to add fans to the charger box to keep everything nice and cool.

Practical:

Adjust-ability of the meanwell. Just how high or low can I safely take the 12v meanwell unit? What about current limiting? How well will the r33 mod work here when current gets up to 30a?

The PSUs do have some pots that enable a SLIGHT adjustment of the voltage, but nothing grand.

So far I've managed to get the PSU up and running by shorting some pins together. However, as soon as I load the unit with a 100w load the voltage drops down to 11.5v. This is unacceptable I believe. I need to figure out how the sense line works. This may be a pain as there are lots of pins to play with.

Fuses on the main charge lead?

FYI

This is going to be used for a 18s lipo pack. (for now at least )
So (12v * 5 PSUs) + the 12v meanwell = 72v
18s * 4.2v = 75.6v
I should have no problem boosting the PSUs and meanwell to get up to 75.6v

1500w (the breaker in my house is the limiting factor here) / 75.6 = 20a - not too shabby. I hope one hour recharge times don't effect the life span of 15c lipo.

 

[Fishmasterdan]

Hopefully these help you find the correct jumper configurations.

https://sites.google.com/site/tjinguytech/my-projects/server-ps
http://www.rcgroups.com/forums/showthread.php?t=1292514

The Server P.S. are the shints...

I did the isolated P.S. trick but used a PVC box. Worked for a while but one ended up burning itself out. If you get a PS that does not have the exact voltage as the other ones the one with a weird voltage will burn up!!.

Dan

 

[auraslip]

I'm starting to think I'll never figure out what ALL of the pins do on these supplies. Maybe that's not a problem, but it still bugs me. I followed this short guide to finding the v+ sense line and v+ current sensing line by connecting v+ out with all the pins and looking for a change in output voltage. I found none! So.... what now?

Well, I've found a way to boost the voltage and it load tested it at 100w for around half an hour with no problems, but I'm not positive what I'm doing exactly.


Load testing with my battery pack discharger


Bottom pin 4th from the right puts out around 11.5v - When connected top pin 5th from the right voltage goes up to near 12v.
Top pin 4th from the right puts out around ~5v - when connected to the top pin on the far left voltage goes up some more. To be fair, it seems any positive voltage connected to the top left pin causes the voltage to go up.


One of those pins on the little white connector of the control board is at ~13.5v - could this be some sort of sense line?


The back side of the "control" board. You can see the three pots. Only one of which seems to do anything.

If it would help you help me figure out the mysteries of this PSU, I can label the voltage of each pin.

 

[auraslip]

I can add the voltages of more pins if it helps anyone.

 

[Hillhater]

so you have 3 big "positive" contacts , but only one small "negative" contact ..???
something tells me you should have a bigger negative terminal fo a 60A PSU. ?

 

[auraslip]

I knew I should of corrected that picture. Top three are +V. Bottom 3 are -V plus the pin closest.

 

[hillzofvalp]

AC can get out on the DC ground wire in some cases.. careful

 

[Hillhater]

I knew I should of corrected that picture. Top three are +V. Bottom 3 are -V plus the pin closest.

Ahh ! OK ..yes, very mis leading that !

 

[auraslip]

AC can get out on the DC ground wire in some cases.. careful

AC ground is floating. Not connected at all.

 

[hillzofvalp]

Some DC grounds are connected to the case...or can become connected to case after a 0 ohm resistor blows

 

[auraslip]

T1 Current sharing bus

T2-6 IC^2 bus M24C32

T7 PIC16F76/73 pin 18 RC7/RX/DT

T8 ON/OFF

T9 Bunch of op amps?

T10 ON/OFF

B1 Bunch of op amps? - control board pin 6

B2 Bunch of op amps? - control board pin 5

B3 Bunch of op amps? - control board pin 4

B4 -V sense on UC3907

B5-B6 null

B7 ~12v from ???

B8-9 null

B10 negative

This took me a bit of time to figure out. Luckily I now have 4 pins that go to a series of op amps on the control board that I haven't yet traced, the ~12v pin, and the -v sense from the UC3907 current/load sharing chip. It looks like I CAN just push some voltage into the current sharing bus and expect the output to rise.

 

[auraslip]

What's this thing on the AC power wires? Do I need it?

I tested the following at 100w for a few minutes:

-V to B4 -V sense on UC3907
B7 ~12v to B1
+V to T1 Current sharing bus

12v

12.4v with adjustment pot cranked up

12.65v with adjustment pot cranked up and +V being fed into current sharing bus.

 

[auraslip]

Well, I load tested one at 200w for half an hour today. It held 13v for the whole time, so I am satisfied enough to start building.



Final design before I start building. The six units I haven't burnt out, isolated by 1/2" wooden spacers, but allowed to "breathe" for better cooling.

Fans on the back of the psus blow air out the back of the main case. fans on the front to help (nesseacary?) -

Room for two s-350-12v mean wells for if I ever upgrade to 220v and 24s lipo.

Panels on the front, but I may just put a turnigy watt meter between the power bank and meanwells. It won't function as a voltmeter at 60v, but it least i'll have an ahmeter.

Fuses? I'm planing one before the on and off switch on the AC side, and then one somewhere on the output side. Somewhere that a fuse would actually be useful (fuses that can break HV DC are rare)

 

[auraslip]

Some times it's easier to do more with less. I didn't actually have enough wood on hand to build what I wanted right now, but what I came up with is simple and will be perfectly fine for at home use.

Just to make sure though - if my cat knocks a screwdriver or something across two of these, they should just shut down automatically like a normal short right? These things are pretty resistant to such errors, but I wonder if because they're in series it'll cause problems?


Quick hour job with scrap wood and hand saw.



Maybe I won't need case fans after all?
These things get pretty hot even at only 200w (what I'll be running them at and only 1/3rd their rated power), and I wonder if part of their design is to radiate heat via conduction through the case of the server they're built to be in.
If so, I might need some cooling soon or later.

Now I just need to mod and burn in test all six of these. Ack. That's gonna be repetitive. But easy.
Then put all six in series and test loaded with the bulbs.
Then get meanwell.
then happy charging

 

[amberwolf]

Many PSUs shutdown by crowbarring (shorting out with an SCR or similar) one or all of the voltage lines, causing the overcurrent limit to trip or the AC-side fuse to blow if there's no overcurrent limit/shutdown.

I am not sure what would happen in the case of series-connected units if one of these did that. The crowbar is only meant for a very short term shortcircuit path, but with series flow, current would keep going thru it once started, and maybe smoke the crowbar circuit.

 

[auraslip]

Many PSUs shutdown by crowbarring (shorting out with an SCR or similar) one or all of the voltage lines, causing the overcurrent limit to trip or the AC-side fuse to blow if there's no overcurrent limit/shutdown.

I am not sure what would happen in the case of series-connected units if one of these did that. The crowbar is only meant for a very short term shortcircuit path, but with series flow, current would keep going thru it once started, and maybe smoke the crowbar circuit.

Just like on a meanwell during an overvoltage condition.

 

[texaspyro]

What's this thing on the AC power wires? Do I need it?

It's a ferrite "bead" common mode choke (inductor) it is used to knock the sharp edges off any switching transients that may get on the wires. They are basically a hack used to get the thing past FCC radio emissions interference testing. In the real world, they don't do anything useful.

 

[auraslip]

It's a ferrite "bead" common mode choke (inductor) it is used to knock the sharp edges off any switching transients that may get on the wires. They are basically a hack used to get the thing past FCC radio emissions interference testing. In the real world, they don't do anything useful.

So I should put one on the ac input of all of the power supplies? I don't want it to interfere with my wifi.



New problem! A fast paced cricket noise coming from the AC side of one of the PSUs. It's either coming from that transformer I'm pointing to, the fets next to it, or a larger fet looking thing behind that heat sink. Some of the low voltage stuff on the output side of the board is working, but I was shocked to see that some of the AC voltage was REALLY high. Like 500v! Scary. I could probably pull replacement parts off the two that I shorted stuff on the output of.

I was pissed too because I didn't test it first and I had just finished modding everything on it. Well I tested all the others, and they work fine so no problem there.

 

[texaspyro]

So I should put one on the ac input of all of the power supplies? I don't want it to interfere with my wifi.

I've never seen one do diddly squat about reducing interference in the real world. They can knock off the edges of the signal enough to pass some arbitrary FCC test value, but if the supply is radiating noise to begin with they won't do enough to help out in the real world.

WiFi runs at microwave freqs. Power supplies switch at near-audio rates. RF interference from them seldom reaches any where near WiFi freqs. It's interference from the CPU and memory clocks/busses that tend to worry WiFi.

 

[auraslip]

SUP BABY
4 of them pushing 12v bulbs in series

All of them power on together. All mods done. I love the click all the relays in them make, slightly out of time, as they power up.
I ended up setting all at the highest voltage I could get them. Between ~12.9v-13.2v, and no I'm not worried about the difference in voltage. You need to worry about that when they're in parallel.


Did 4 units today - coffee and the punk rock station on http://grooveshark.com

Stripping the lil wires was time consuming, but much quicker if you assembly line it.

Made a little drawing so I wouldn't forget.

Bad picture, but this is how the main charge leads get on there. Sort of a pain. Lots of solder and a BIG soldering gun required.

For the board to fit with the bottom charge lead, I had to cut off a small part of the bottom of the case.

More pictures tomorrow.
All I need now is to install the power switch and the power jack.
Then a meanwell + battery pack next month.

I'm thinking of doing a nice write up of this project on my site. Might get some press on the gadget blog sites. This project is fantastic from a time/value perspective. It'll probably be something that is very useful for larger EV conversions. Those big chargers are $$$, and this one is only $100. And it is usable with any pack or whatever you want to use it for; from a ping pack up to a full on motorcycle sized pack. If it works, it'll probably be the last charger i need.

 

[ZOMGVTEK]

I got 8 DPS-600PB for <$100 shipped, which are good for just nearly 5KW.

I just need to isolate the case from DC ground, put Powerpoles on the outs, and they're good to go. These guys put off some seriously clean and reliable power. I plan on running them in series with a 60A Sorenson to make a 0-150V 50A supply for <$300.

How are these supplies working out for you?

 

[auraslip]

Damn dude. You win.

Supplies are sitting, waiting for the meanwell and battery pack. Should have it running (as well as my new bike) in the next month.

 

[ZOMGVTEK]

99.8V @ 47A with all of them in series. They can be paralleled without complaint as well.
I'm really happy with these supplies. The voltage is somewhat adjustable, but its basically just a 12.45V supply.
I isolated the DC ground from the chassis, all of them have earth ground connected and are in physical contact with each other. The tape just holds them together.

Still, they can be found for $10-15 each shipped. The 48V 3KW version is a better bang for the buck, but this unit will do 12V increments.
All you need is a 12+V 50+A variable supply in series, and you have a 0-112V 50A variable.
If you can find a good deal on a 50V+ 50A+ variable, the 48V 3KW HP Server Supplies will do the job on the cheap.

 

[dennyt]

ZOMGVTEK - very nice! Can you elaborate on how you isolated the DC negative from the chassis? I picked up a few Dell DPS-500CB's (12V 41A) and I'm getting ready to assemble them in series. I like the idea of keeping the cases grounded and isolating the DC negative.