Results 1 to 8 of 8

Thread: Dual HP DPS-1200 PSU's with iMonitor activated

  1. #1
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Dual HP DPS-1200 PSU's with iMonitor activated

    I've got a idea bugging me that I imagine someone may have already solved, or else I'm just an eccentric who overanalyzes everything to a point ad nauseum and need to just let it go! :-)

    A while back I "stumbled" into the HP Server PSU alternative 12v power supplies and think its a good alternative to the open frame Meanwells, etc.. Especially since it is delivering 2x the power for the same purchase price.

    One of the features I think is quite beneficial is the load leveling/current sharing that can occur between paired/parallel PSU's by tying Pin 34's together. At 115vac input thats 150 amps@~12vdc. At 230 VAC thats 200 amps! In a ~ 10"L x 4"W x 4" D package!

    There's even a 12vdc standby output at around 2.5 amps per PSU to keep power to any low power processors.

    I'm no PCB designer, but been taking a crash course in Circuit Maker to try my hand at it anyhow. I'm a bit outmatched at this point, but will continue to tinker away at it, and wondered if anyone had maybe already tackled this particular design, or would be interested in offering any advice.

    The amphenol card edge connectors are readily available, and the chinese alternatives are even more affordable, so that's not the hold up.

    Part of this concept has already been tackled, but for purpose of the 24vdc "float" mod as the 12vdc "mega power" demand is limited.

    The idea is as follows:

    *Stack 2 psu's using a 3d printed holder/sleeve/mount - this also locks down the card edge C/L's for the board to engage.
    *This would requires an STL but there are already some out there that are probably adequate to the task

    *Have a single PCB board with two card edge connectors on the established C/L this PCB face would be perpendicular to the PSU card edge (I think its a 0 degree connector)

    *The PCB real estate aligned with the +12vdc pins is shared by both PSU's. Same for the GND, no reason to have different fields/copper pours for each as the whole purpose is to combine the output.

    *The other pins along the far side would be used to connect the IMONITOR feature, start switches and any other features that seem reasonable, set the pmbus addresses, etc...


    There's not enough room to put fuses or circuit breakers without stretching the CL, but if it was just a short jump into a fused or CB distribution block, that should be fairly safe.

    Think car audio or welding 0-2 GA cables attached to a connector bus staple with lugs, through PCB or SM with lots of vias to assist with current flow (maybe a 2oz board as well?).

    The PCB vertical arrangement doesn't add much to the OAL since it'll likely only add ~3/4" after sliding over the card edges.

    ALL on the same potential, so no hassle trying to make sure wires between typical 30A psu's don't accidentally criss cross....



    Maybe $20-30 in board parts overall depending on the card edge connector mostly (2x @ $3-7) and the lugs (4x @ $2/each) a few resistors and a switch, plus PCB which could potentially stay below the 100mmx100mm package size.


    Basically, for under $100 you could have 1800-2400 watts distributed out of a single ammo box..


    crazy, isn't it...?


    SC

    "Dum spiro, spero"

    General Parts Involved
    2x HP DPS-1200 (HSTNS-PD11 in my case) Server PSU's or equivalent (~$25Each delivered)
    1x 100mm x 60mm PCB ($No Clue$)
    2x Card Edge Connector - Amphenol 10046971-001LF or equivalent (64 pin 2.54 pitch card edge connector) (~$4-5/ea at Newark for Amphenol-less for chinese version)
    1x switch - C&K Components JS202011SCQN or similar (Taken from PB16 build list) (<$1)
    2x 21k ohm resistor ($0.50 each)
    4x 100 amp rated connectors (2 per 12vdc, 2 per GND) (<$2 each)
    12x 16-18 gauge jumpers (Free DIY)

    1x TBC...


    Running total==> $50+PCB?+$10+$1+$1+$8+$0= $70+PCB?+TBC
    Last edited by SuperChuck; 01-23-2021 at 12:33 AM.

  2. #2
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Quick sketch formed from an stl I found.


    dps-1200 dimensions.jpg

  3. #3
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Showing PCB(Blue) and card edge connector (purple) arrangement


    close up of PCB arrangement.jpg

    A little more detail using 3d model...

    PCB DPS-1200x2.jpg


    PCB DPS-1200x2 side view.jpg


    With 2 ga lugs

    PCB DPS-1200x2 with lugs.jpg
    Last edited by SuperChuck; 01-22-2021 at 11:39 AM.

  4. #4
    Join Date
    Dec 2019
    Posts
    330
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    I've thought about doing something similar, it just hasn't become necessary yet. Last year was my first year, so that might change, but I found that I have a tendency to keep the pixels per port lower, so I haven't even needed to power inject yet.
    However, I still think it's a great idea! So I'd like to keep track of your progress. Also, I have ready that 4 of these can be connected in parallel!

  5. #5
    Join Date
    Dec 2011
    Location
    UK S80 postcode
    Posts
    1,519
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Hi, this is an interesting idea. Given that HP are a reputable company there is little doubt that the track width on the PSU output and a good quality edge connector is capable of passing the required current. Where I think you will need to be carful is the copper thickness on your PCB and how you take power off the PCB. 75/100A from a single supply is quite a bit and going to need some research to get it working first time. Heat is going to be the most likely issue, I’m sure you have come across burnt terminal on power distribution equipment.

    There are ‘PCB Trace Width Calculators’ online which can help you with designs. I used this one to get the information below: https://www.4pcb.com/trace-width-calculator.html

    From photos of the HP DPS-1200 it looks like the output copper traces are 13 X 0.1” wide (1.3” or 1,300mil), double sided (1,300mils x 2 = 2,600mil or 2.60e+3mil in scientific notation) and used with a 240V supply will need to carry 100A.

    If you feed those figures in to the yellow inputs/optional inputs part of the calculator:

    Current: ‘100’ Amps.
    Thickness: ‘2’ oz/ft^2.
    Temperature Rise: ‘10’ Deg C (more later).
    Ambient Temperature: ‘25’ Deg C (more later).
    Trace length: The distance between the centre of the two PSU is 1.6” so if the output connector is at the midpoint the ‘trace length’ on the PCB is 1.6” / 2 = ‘0.8” ’.

    This should give you some results in the green fields, I’m assuming you are only thinking of a double sided PCB with 2oz/ft^2 copper so there will only be ‘External Layers in Air’:
    Required Trace Width: ‘3.39e+3’ mil (3,390mil or 3.39”).
    Don’t worry to much about the ‘Resistance’, ‘Voltage Drop’ or ‘Power Loss’ for now.

    What does that all mean:

    If you are going to pass 100A through 2oz/ft^2 copper traces over distance of 0.8” in an environment with an ambient temperature of 25 Deg C and you do not want the temperature of the PCB to rise by more than 10 Deg C you will need a track width of 3,390 mil.
    HP only have a track width of 2,600mil so they use a thicker copper or accept a hight temperature rise. Have a play with the values (one at a time is best) and see what effect this has.

    I’m guessing 2oz/ft^2 copper as its cost effective and a temperature rise of 16 Deg C above Ambient. 25 Deg C ambient + 16 Deg C PCB temperature rise above Ambient = 41 Deg C for the PCB as a continuous full load running temperature. This now becomes our new Ambient temperature for the rest of the calculations.

    The contacts of the ‘Edge connector’ you use will also have a resistance and passing that sort of current will generate some heat, this in turn will by the nature of the fact that it is soldered to the PCB copper also raise the effective Ambient temperature of the PCB.


    Now to a part that is not so easy to explain or understand. I’ll use an analogy to explain this as it’s easier than the maths (values have been picked for ease of explanation and are not meant to be technically accurate just convey a point).

    The current comes out of the PSU through the edge connector onto the copper tracks of the PCB which we have looked at and should work thus far. Now the current has to exit the PCB through a connector or connectors. Let’s assume a single connector rated to pass 100A with a diameter of 1/2 ” (500mil) .

    Analogy: Think of the Sun as the 100A current and based on the previous calculations if focused through a magnifying glass to form a circle of 2.6” (2,600mil) on the PCB, that will rase the temperature of that area by 16 Deg C.
    Now imagen shining the sun on your hand with the same 2.6” diameter circle and feeling the warmth. Then reduce the 2.6” circle down to a 1/2” circle. More energy is focused into a smaller area and is going to generate more heat (it may become painful).
    A long winded way of saying the power delivered from the HP PSU is through 26 contacts per polarity spaced over 2.6” of copper. To try and take that power through a small area of the PCB will generate heat in that area and may if unaccounted for cause problems.

    Basically, if it was me I would work out the optimum number of connectors to fit in the 1.3” (double sided) space available that can carry the 150/200A you can draw from the PCB and spread the load draw as evenly across the track width as possible. This will keep localised heating down and reduce the risk of damage due to over heating or the long term effect of expansion and contraction.

    On a side note: It looks like the PCB may partly block the airway of the lower of the PSU’s.

    These are just thoughts and not intended to stop you trying something. If your show doesn’t draw full power for any significant time many of these issues reduce, but if your controller freezes on full white then these issues may become more of an issue. Its great to see someone with a project early in the year, do some max load or destructive testing and then you’ll have all the answers before show time.


    P.S. Having read through this before posting it’s not ideal, very Tec and a bit to blunt, there was quite a bit to cover. Please fell free to reply with questions on each part and I can provide links and others will chip in with advice.

  6. #6
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Barnabybear,

    Thanks for the technical approach on the current flow issues, I've glimpsed the calculators early on and did not get to the level of detail you've shared, but realized it was a significant part of the challenge. I agree/understand the assessment and the assumptions you've made to arrive at them.

    I think a few boundary conditions are definitely required to model the system, and your values are good;

    200A - Total Amps
    2x - # of connection pairs(12vdc/GND)
    100A - Max current per pair

    The reason that I've selected two connections pairs is that as you increase connections, and decrease amperage, the space efficiency goes away very quickly.. I could probably make 3x @ 65 amps or 4x @ 50A each work, and those options are on the table(3 is my preferred max). 5 @ 40A or 8 @ 25A does disperse the amperage over more board but also increases number of conductors, landing points, etc... Still not a deal breaker, but was just not the direction my mind first wandered.

    I have gravitated to this particular lug after first finding the ones in the graphics above. and have a few reasons that this helps to mitigate a few of the concerns above. It is solid copper, has a 95A rating with copper cables, the staples mount through hole, but the copper body can be SMT soldered as well.. Cross Sectional path to the cabling is increased significantly. It is a little smaller than the ones shown above, so 3 could be used if needed - although a few more factors make me lean to 2 per potential.

    copper lug.jpg

    One other item - the through hole staple could be replaced with an SMT staple which is basically a SMT mount shunt or jumper. (shown in second picture)

    shunt jumper.jpg

    The surface mount shunt/staple is 0.91" long (1.6"-0.91"=0.69"/2= 0.345") and significantly reduces/shortens the path solely in the PCB.

    This also increases the cross-sectional flow path to the lug reduce heating effect from the bottleneck. Combined with soldering the lug to the PCB, and we have a significant improvement.

    PCB Path length drops !!
    Cross-sectional area goes up !!

    smt lug.jpg

    If this area of the board was not masked, or had pads for multiple lugs(3-4 per potential), the assembler could make the ultimate choice of how many they wanted to install. I believe that we would need to provide a lot of VIAs to allow for the one-sided PCB mount, but that's easy to design in. The shunts/jumpers could also be used on the opposite side to reduce the burden on the PCB.

    Thanks for your feedback, I enjoy collaboratively attacking and solving problems!

    Regards,

    SC

  7. #7
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Per the previous discussion, the lugs were upgraded to the SMT mounts and jumpers were added for visualization.

    After looking at off the shelf SMT and Through Hole Jumpers, the simplest jumper we can probably design for is some 18-16 GA solid conductor wire that we can easily scavenge from our benches or scrap bin.

    I know every time I change out fluorescent lamp ballasts to LED conversion there's always scrap wire afterwards that would be perfect for jumpers. Each one could handle 5 amps or more point-to-point without ever breaking a sweat considering the Amphenol pins are rated 7A.

    If you like the store bought jumpers, you can always buy them instead, but we could design in 10-12 or more jumper positions and completely change the characteristics of the PCB resistance.

    SMT w-jumpers.jpg


    FURTHER DISCUSSION

    We already have some assumptions for board max design values.

    POINT
    200A is based on 100% load with 230VAC input
    2400 Watts at 94% efficiency = ~2560 watts to the PSU
    ~11.1 amps @230 VAC

    COUNTERPOINT
    115 VAC is de-rated and only yields 150 amps
    1800 Watts @ ~ 94% efficiency = ~1920 Watts to the PSU
    ~16.7 amps @115 VAC

    The best way to run this system would be at 230VAC due to the extension cord load, and doing so means it has the potential to deliver the full 200 amps@12vdc.


    In practice....

    Unless we monitored the current, we'd likely estimate the allowable loading, and limit the demand on the system after a certain number of pixels based on accepted rules of thumb based on full load conditions(which provide for a generous safety factor). Also, current demands are dynamic, hardly ever reaching actual or theoretical peak demand.

    Even if we DID design using an ammeter to set loading, most people would leave a reasonable amount of "leftover" capacity and not design down to the last milliampere.


    These same conditions exist for ANY capacity PSU, Meanwell, HP Server, etc. etc. so we are not introducing a new concept, only recognizing its impact.

    People are likely buying excess capacity due to rules of thumb or redundancy, or simplicity.

    What this does create is a "scale inefficiency"(my term). The greater percentage an average base load is compared to the total load, the more likely the total capacity is less efficiently utilized.

    If the average incremental load is 1A, then we can easily sneak up to the 27 amp range on the 30A PSU or even closer, which improves utilization. If the average block is 8A, then you may likely limit the load to 3 blocks/24A, because that 4th block exceeds 30A based on the assumptions. If your circuits are 16A, you will only attach 1, etc.

    With a base block of 150A-200A you can access that "lost" capacity and potentially save money. 8x 16A blocks @150 compared the example above which requires 8x 30A PSU's.


    All of this philosophical discussion to say, most likely even a 200Amp supply will only run at an average loading of 75-85% because we design with the safety margin built in and the assumptions cover worst-case.

    BUT... the smaller capacity PSUs may very well be forced to operate at even lower utilization 60-75% due to the "block ratio" issue discussed above...


    Sorry, I really got off on a tangent there. :-)

    SC
    Last edited by SuperChuck; 01-22-2021 at 11:47 PM.

  8. #8
    Join Date
    Jan 2021
    Posts
    6
    Post Thanks / Like

    Default Re: Dual HP DPS-1200 PSU's with iMonitor activated

    Glycerine,

    I think the communication bus and the IMONITOR circuit allow for up to 4 PSU's in parallel.

    In series, which is where the Lithium Ion charging users are playing with the "floating" mod - they do not have to communicate, as they are being "stacked" to increase the voltage.

    "Load sharing" really becomes load contribution, as their current contribution matches the contributing PSUs.

    Their Power contribution in theory should be the same for each PSU if they are all set to the same voltage. If not, their Power Contribution is directly proportional to their voltage contribution to the stack.. (V1/Vnew = P1/Pnew - see below)

    V1+V2+V3+V4=Vnew

    I1=I2=I3=I4=Inew

    P1=(I1*V1)..... Pnew=(Inew*Vnew)


    Quote Originally Posted by Glycerine View Post
    I've thought about doing something similar, it just hasn't become necessary yet. Last year was my first year, so that might change, but I found that I have a tendency to keep the pixels per port lower, so I haven't even needed to power inject yet.
    However, I still think it's a great idea! So I'd like to keep track of your progress. Also, I have ready that 4 of these can be connected in parallel!
    Last edited by SuperChuck; 01-23-2021 at 12:18 AM.

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •