Testing the HP ESP-135 Server supply and isolation modding.

Joined
Oct 28, 2008
Messages
14,539
Location
Manhattan Beach, CA, USA
So, the little HP ESP-135 server supplies are dirt cheap, like $15-25 shipped sort of cheap. They must have a hundred bucks or more of materials and labor in each one, but the datacenter industry keeps the market flooded with used ones.


You can power an RC charger with them, or use them in series with a lab power supply or a CC-mode modified meanwell to be a battery charger. Putting them in series with something like a 20v 50amp lab power supply boosts the voltage so you can use it to charge something like a 18S LiPo pack for example. You would put 5 of these 12.65v output power supplies in series to give you 63.25v, then set the 20v 50amp lab supply to 11.25v and 50amps, and it would charge your pack at 50amps up to 74.5v (4.15v/cell), the server supplies would be a fixed voltage, and the lab supplies voltage would provide the current limiting and slowly climb to 11.25v as the pack charges.

This type of charging setup is fairly advanced, definitely something to try at your own risk after you've got plenty of experience and are comfortable with what you're doing. Otherwise just use them to power something like an RC charger, and let it handle the charging control.

Here is a video of the output being tested. They all seem to trip-off right about 56.x amps, so as long as you're not pulling over that amount of current, they just stay solidly fixed at about 12.4-12.65v depending on load.

[youtube]RHvGg64onnc[/youtube]

The power supply is a fully isolated type of architecture, but they intentionally tied the DC ground output to the chassis and AC ground in 2 places. Fortunately, these are pretty easy to remove from the DC side, which retains the AC chassis grounding, and allows us to run them in series while still retaining the AC chassis ground.

Remove the various chassis screws holding the sheet metal panels on and remove the top panel.

IMG_20120527_001031.jpg


Pry the insulation layer's sticky goo adhesive off the other panel so it can be slid back from it's attachment slots. It will still be tied by the fairly short fan wires, so be gentle.
IMG_20120527_001051.jpg


Remove that screw I'm pointing to. (all ready removed in picture)

IMG_20120527_001131.jpg


With that screw out, drill that pressed in nut out of the sheet metal from the back until the little stand-off body is free.

IMG_20120527_001342.jpg


Pry the little stand-off nut you drilled out from under the board so that corner of the PCB is floating not contacting the chassis.

IMG_20120527_001447.jpg


Drill a couple of holes in the sheet metal next to the fan sized large enough to fit your wires through. I used silicon 6awg wire (because it was laying around), I think 8awg would be fine and you wouldn't need as large of holes.

IMG_20120527_002258.jpg


This is what the back of the supplies connector looks like. See that mess of pins in the middle? You need to connect just 3 of them together, and make sure the other pins are shorting to anything else.

IMG_20120527_003221.jpg


Now the three pins are bent so they are touching, and ready for a bit of solder. It's the lower left corner pin, and the middle two pins on the right side that need to be shorted together for the power supple to believe it's in a server and needs to power-on. Thanks to some RC groups guys for finding out which pins to connect to power these up.

IMG_20120527_003841.jpg


A few supplies with the pins bent and soldered.

IMG_20120527_004020.jpg



The PCB has 2 jumpers right where the tip of my dikes are pointing, you gotta snip them both and ensure you have no connection between those two sides anymore. This isolates the other point where DC was connected to the chassis.

IMG_20120527_004146.jpg


I crimped some ring terminals on and bent them to route decently from those two nice stand-offs. You can see my screw drive on one of the stand off screws you want to bolt one terminal on, the other is just down and to the left by a cm or two. These are the positive and negative output terminals of the power supply.

IMG_20120527_004730.jpg


All bolted up.

IMG_20120527_005158.jpg


I slid some shrink around the wires in the area where they pass through the holes drilled in the sheet metal, just to help avoid it cutting the insulation and shorting. If you wanted to do it right, a grommet would be a good idea.
IMG_20120527_010039.jpg


That's about it. It took me about 3hours of fooling around to do the first one and figure out how to isolate them and how to cable to them. Then doing 3 more of them took about 2hours of medium-easy work.

I would rate it about a 3 out of 10 on a DIY difficult scale. I would rate a 9.5 out of 10 on a value in power supply vs money spent scale. :)

-Luke
 
Can't find a reliable source on this, does this accept 220VAC?

Other than that, great guide :)
 
Awesome, thanks for the guide and photos. DC isolation sounds like the way to go.

I have 8-10 of them sitting here waiting to be pimped. The first 2 I put in series with just one AC plug's ground pin removed and have been using them to power an iCharger 3010b for months now up to 1kW output.

The rest I'd like to use in series with an adjustable current lab power supply. Ideally 0-12v or so and 0-60A adjustable, though the closest thing I've found with a somewhat sane price locally is 0-40A:
http://www.electusdistribution.com.au/productView.asp?ID=13800

An RC guy did a write up about them here too for anyone wanting to see more photos/diagrams:
https://sites.google.com/site/tjinguytech/my-projects/HP47A
 
That's interesting your supply would fire up under that high of a load. I've had them sometimes not start up under anything above about 20A, and in this series rig they are even more picky. One or two never start up under high loads. They're usually not the same supplies.

http://farm7.static.flickr.com/6211/6289690286_3f554993f5_b.jpg

http://farm7.static.flickr.com/6158/6199798968_f4d1f45d04_b.jpg

I bent and soldered the 3 pins, smashed 10 AWG between the output terminals, soldered it, and covered it with glue. I isolated the DC off the chassis on all of those supplies, and the tape actually holds the brick in electrical contact. I simply cut the PCB around the standoff with the pad that goes to DC -, and wedged a screwdriver into the jumpers to break them. It's really fast, I finished that brick in less than an hour.

On the RC forums, theres details for how to mod them to pull up the output voltage, or pull down the fan speed. I slow the fans down a bit on most of these. I have the fan really slow on the one I use on my iCharger, the voltage around 13.4V, and it's been fine at 200W or so with weeks straight of idling to cycle packs.
 
Thanks Luke for the pics. These are also known as DPS-600PB power supplies if youre searching on ebay for the lowest price. There are literall tons on there and super cheap. I got two shipped last week for 20 bucks from savemyserver

Instead of spaded wires you can add banana connectors (which allows easy series/para reconfiguation for other uses) or just solder to the outside terminals. Also a neat tip I picked up is to take off the end handles and use them to tie two units together on top (after youve lifted dc on one!). There is a mod to slow down fan speed but why reduce the life of the components. You can also add a switch to turn on dc but that doesnt disconnect ac so your better off with a cheapo power strip
 
Whats the digital thingy with the dial on it worth and were do we get it? I'm asking because I have 2 50v 3000 watt server supplies. But I think you said you need to limit the current? And adjustable current would be cool....
 
Stork- Yes, 240vac input is fine. I think the range says 100vac to 290vac.

Yes, these ones seem to start fine loaded, I did that test about a dozen times on all 4 of them with no faults .

Arlo- The big white box is called a DC load. This one is made by BK precision and can do 240amps max and 160v Max, but no combo of amps and volts that exceeds 2400w (its heat sink dissipation limit).
 
Also, though it would work to use a DC load for current regulation in making a battery charger work, it would be super bad efficiency and extremely expensive. That load was $8000, and a lab supply to do the same thing would be about a grand, and way more efficient, as it just makes the power you want rather than making more than you wanted and bleeding off the rest as heat.
 
I didn't isolate mine because it was such a tight fit that I was scared it would short out when closed. Also I was lazy and didn't want to do all 6 of them. Instead I just put them in a wooden frame.
P1020499.JPG

Kinda ghetto but works. just don't try not to drop metal tools on them :lol:

anyways, here is the guide to using these I wrote for my site. should be helpful https://sites.google.com/site/shelbyelectro/batteries/batteries/server-psu-charger
 
Thanks Luke, I have been meaning to do this but the other sites weren't too clear on how.
Instead of bending and soldering the little pins, I push some JXT SM's, one and two pins[cut off from unused controller leads]down over the three pins and join them together.
 
A couple of photos of my units. PCB snipped instead of standoffs drilled out, DC jumper wires broken apart. I also snipped the top of the plastic cover apart to make it easier to add power wiring to the front solder up the little pins.

View attachment 1
esp-135 mod 2 s.jpg

Comments/suggestions welcome. Cheers!
 
Nice guide, these are nice units and it's nice that you can run the outputs in series with a little modding.
Teh Stork said:
Can't find a reliable source on this, does this accept 220VAC?

Other than that, great guide :)
Hi, I'm using this exact PSU to power a 12v RC li=po charger, and it works great from 230VAC. :)

I put a 31ohm resistor on the fan to make it quieter, as I'm never gonna pull more than 20A from this PS.
 
I'm powering on with creating a scalable charger using these power supplies in series with a lab supply with adjustable voltage and current.

Here's where I'm at, just starting testing:

powertech_and_2_esp135_in_series.JPG

The image shows the $330 power supply I've just bought which has 0-16v and 0-40A both adjustable. On the Cycle Analyst, you can see 2 of the ESP-135 are in series with the Powertech unit adding ~12.5v each. The supplies themselves are not visible in the photo. Next step is to add big terminals at the back and test at higher current, and mod more of the ESP-135's to run in series.

The lab supply is working perfectly as expected in my limited testing, connected to a 10S Li-Mn battery. I can charge the pack with the lab supply in either CV or CC.
 
Well, this thread's gone a little quite but I'm still powering ahead to make my 75-90v 0-40A charger.

Here's what's in the photo:
1 x Lab power supply 0-16v 0-40A (640W max charge power)
6 x ESP-135 (500W charge power @ 40A limited by lab supply)
2 x power boards with 8 C13 (10A) inputs and a C20 (16A) output
7 x C13 to C14 short leads

photo 2 (6).JPG
photo 1 (4)s.jpg

Each power board has 8 inputs, so I could have used just one, but having a total of 3.6kW of charging power is too much for a single common Aussie 240v 10A outlet. Instead, by splitting into 2 powerboards, I can connect to 2 wall sockets and have plenty of headroom.

I have plenty of work to do for a rainy day - change C20 plugs to standard Aussie 10A plugs (AS 3112 or Type I) & make the DC series wiring for the front side. I'd also like to package everything into a nice little road case.
 
Does anyone have experience putting these in parallel? Looks like everyone is doing in series, or in parallel and series with a current limiting supply.

I need to put three of them in ||, no series.

I realize I could use ballast resistors or schottky diodes, but I don't have either beefy schottky or beefy low resistance resistors on hand. I will procure them, if I have to, but that takes time and $$.

I was hoping...can I get away with just assuming they're sort themselves out? I figure they're high spec server supplies, and the same model, so I might be ok.


I intend to run an etek brushed motor coupled to a 24" drum sander [for foam] at 12v....I hope they can tolerate the current surge because I don't have any soft start except whatever inductance is in the motor (I'll use whatever very high amp solenoid or knife switch that I can find.)

Just throwing this out, in case there's some stupid alarm going off and someone with magic ears will let me know what I can't myself hear.

(I'll keep going through http://www.rcgroups.com/forums/showthread.php?t=1292514 in case I find more info). One of my solutions will be to use http://www.rcgroups.com/forums/showthread.php?t=1005309 instead. 100A supply. But, it's a little shy of the power i want, especially during surge.
 
They do accept a short circuit condition for about a half second before they trip.

They also start up under full load (50amps) just fine, but they won't start up shorted.

I think with 3 or 4 of them, and about 10ft of 6awg cable for each one to act as a current balancing resistor, I bet it will work for you.
 
Ah! Not bad, I should have thought of using some cable as a resistor.

If I use a thinner gauge cable, do you think it simply won't tolerate the amperage? 6awg seems excessive on each line if they're only running 50amps each, and I don't need such a long cable. Plus, it'll be a lot more expensive.

X-section of 6awg = 13.3mm^2
X-section of 8awg = 8.37mm^2

So, if you're figuring 10ft of 6awg, I guess 6ft of 8awg will work. I think 8awg is capable of tolerating 50amps, at least I don't expect such high constant load. Then, I could pick up the wire from hobbyking. And...were you thinking that I'd need the long wire on each wire side [Positive and Negative] (i.e, 20ft each PS), or just 10ft total per PS?

If it's just 10ft total per PS, 10ft of 6awg is 4mOhm (according to http://www.cirris.com/testing/resistance/wire.html?) and good for .2v drop @50amps.
[I found the resistance calculator after I did the stupid cross section babble above, but they both check themselves.)

I wonder if it would be useless or not to loop that wire around something ferrite to make a bit of an inductor. That would...reduce the current surge? I A) don't have any practical inductor experience outside of physics 101, and B) have no idea if that would have a meaningful impact.

If it's 6ft of 8awg per power supply, and I use 3, it'll be something like $20+ shipping from HK. If I need to use 6awg and 10ft, then it'll be more like $70-80 of cabling (~2x the length, 2x of 8awgs ~=6awg.).

Thanks, Monsieur, for the feedback.

I'm thinking especially if I can get away with the 8awg cable, I'll pick up 4 of the power supplies. They're ~$15 on ebay, or I could just buy 10 of them for $78.
 
You can always try 8awg, and I bet it will work fine for you. I've personally just got to the point where I like to put some margin in my cabling, but if you're looking for a budget solution, I think 8awg will work.
 
I finally got around to doing this to some supplies and thought I'd share a couple of things based on my experience:

I screwed around a bit trying to see if it made sense to attach the output leads somewhere else, like removing the "hot-swap" connector and soldering them somewhere, or something, but I ended up deciding it made most sense to do it just like Luke recommended above. I used some 6ga / #10 stud terminals that were a bit too big, I had to remove some material to make them fit on the terminals. I used an angle grinder.

This is what the terminals looked like before installation:
IMG_20121027_114413.jpg

and installed:
IMG_20121027_112245.jpg

IMG_20121027_113448.jpg


I initially didn't like the idea of drilling out the stand-off out of the case, and tried doing it like voice-coils above- but I ended up cutting through a small trace that runs very close to that pad. I tried drilling out the stand-off instead and found it to work out really well, the standoff is aluminum and is very easy to drill out, works out really clean and tidy.

The step-drill that Luke shows in his pics is the best tool I'm aware of for cutting holes in the thin sheet metal case- I don't have a step drill here at home so I decided to do it differently- I have some grommets I threw into a mcmaster order a while back and I just used the angle grinder to notch the case so I could install the grommets. The 1/4" ID 3/8" OD grommet fits very nicely around the 8ga HK wire. There's much less chance of bending the case up with the abrasive cutter than with a drill bit, in general.

IMG_20121027_112140.jpg


Other than those things I did it just like Luke wrote above, and agree that it's pretty easy and fun!

IMG_20121027_122330.jpg


Now for the punch line: I was planning to use four of these and a Mean Well RSP-1000-24 as a ~3kw charger for my big 18s/40AH hobby-king battery pack, powering the power supplies from my "prius generator"- the outlet I installed on the HV battery in my prius- (170-240vdc ish)... most SMPS work fine from this power source and the prius makes a really nice quiet, clean, mobile generator.

These ESP-135 power supplies do not appear to work on DC power - They power up similar to how they do before you solder the enable pins together- the fan spins slow but the main power supply won't kick in. Too bad. I have tested several other power supplies on this HV DC with good results: Mean Well HLG and RSP, a Lenovo laptop charger, some Mean Well S series supplies (with the dual voltage switch set to 240), even a DeltaQ charger all seem to work fine.
 
Aww man! Thats brutal and quite a suprise that they dont turn on from DC! D'oh!

Excellent guide for how to mod them though. Somebody should do one for the 51.5vdc 57amp 3kW HP supplies and show what to drill/snip to isolate them.
 
liveforphysics said:
Excellent guide for how to mod them though. Somebody should do one for the 51.5vdc 57amp 3kW HP supplies and show what to drill/snip to isolate them.
Ill do it. I forgot I have 2 of those.
 
I followed your directions and had a good result; now my hyperion is up and running, dumping out a nice 550w.. :mrgreen: thanks Luke. But i made one alteration to your recipe that is worth consideration.

The main board is very heavy and packed with stuff. I was worried about it warping over time with one of the standoffs gone, and one corner loose.

So i took the whole thing apart and shaved off the green coating + trace on the top and bottom of the two front posts to make them non-conductive.

This lowers the front of the board maybe 0.05mm or so; no big deal, and the board should be just as tough as it was from the factory if exposed to vibration / the elements / etc.

Just something to consider.
 

Attachments

  • hp-power-supp.jpg
    hp-power-supp.jpg
    51.3 KB · Views: 9,494
acuteaero said:
These ESP-135 power supplies do not appear to work on DC power - They power up similar to how they do before you solder the enable pins together- the fan spins slow but the main power supply won't kick in.
Does it matter if you bypass the entire AC-input section (MOVs, rectifiers), putting your DC input power directly to the capacitors on the input side?

Because that *should* work on just about any SMPS I can think of. It *does* work on all the ones I have tried that with--there are some laptop supplies and such that didnt' work with DC on the AC input prongs that I did not go further with, but some of the regular SMPS type PC PSUs and similar that didnt' work with DC on the AC input prongs did work when I put it right on the caps, in the correct polarity.
 
Sometimes they have a relay in series with a leg of input power, and the relay is controlled by the power factor correction stage. In those PS' s bypassing the PFC stage relay (which does nothing when running DC), might work.
 
Back
Top