Server Power Supply Interface Board - for standalone miners and GPU rigs

[supercomputer987]

Sidehack or others that may know the answer.  I currently own several of your boards and they work great.  I am in the process of attaching several boards together to power my equipment more efficiently.  The kits I have received have spade connectors or whatever the crimp on connectors are called (these are attached to the opposite end of the wires where the 6-Pin connectors are).  I was looking for these spade connectors (because they fit perfectly) to build the connecting wires between boards.  Could you please provide me with any details regarding the size/style of the spade connectors or possibly where I could buy them from?  Also, do you have a recommended wire gauge to use when attaching boards together as shown in http://www.gekkoscience.com/products/D750_3way.html?

[1715255410]

1715255410

[1715255410]

1715255410

[1715255410]

1715255410

[1715255410]

1715255410

[sidehack]

I've made jumpers out of three lines of 16AWG, and I've hardwired boards with two lines of 12AWG. It really depends on the worst-case current loading for if a supply goes down. Say you're pulling 1500W from three supplies, fairly balanced wiring, drawing 500W per board. If an end supply goes down and the other two supplies give the full 1500W, you're now pushing 500W into the end board through the bus ties. At 12V that's 42A, which two 12AWG lines can handle with no trouble at all. If you go that route, it's probably technically better to put each bus tie to a different screw terminal, since they're only rated for 20A each.

The spades we use are sized for 10-12AWG wire and a #6 stud. I think the outside measure is about 0.32", I forget exactly but it's a pretty standard spade width for a #6 stud.

[supercomputer987]

Sidehack thank you so much for the detailed response.  I will keep all of that in mind and give it a try.

One more quick question.  Can the common current-share line (and similar) can be wired together with jumper cables (I think generally are 20 gauge wires) common in most embedded electronics development supplies?

[sidehack]

Yeah it's a signal line, very low current. 20AWG wire would probably be more than enough.

[wpgdeez]

I bought a set of 3 boards and on one of them the on\off switch is seized and I cant switch it on. Is there a way that I can bypass this switch?

[klondike_bar]

I bought a set of 3 boards and on one of them the on\off switch is seized and I cant switch it on. Is there a way that I can bypass this switch?

short the pins/solder points for it?

[sidehack]

That's no good. Probably got squished in shipping or something. An easy workaround is to jumper the EON and 3V3 pins on the header. Even if you solder the jumper, you can still use DIP switch 3 to turn it on and off.

[Scoobypup]

Sidehack, how goes the work on the boards for the DPS-2000BB?
Is everything still going as planned for them? I can't wait to pick up a few when they're ready.

[sidehack]

We've had some delays with the PCB design, partly because of other projects and partly because of the back-and-forth required for peer reviews - which become more difficult when there's only two people to look at it and both of them are fairly constantly busy. This PCB is a fair bit more complex than the D750 boards, by merit of the built-in fan controller circuit and some required signal inversions, with all the surface-mount components on one side - plus we're trying to design it with heavy internal traces that won't require the external bus wiring. That's somewhat new territory which merited writing up some thermal dissipation code to help out, and we're talking with a different board house than with the D750, an outfit that specializes in high-power traces. We sent the latest board revision (I believe V0.7) to them this afternoon to check out, hopefully we'll be able to place an order and have a short batch back within 7-10 days so we can assemble and start load-testing them. I spent today working on base designs for a 3KW adjustable dummy load so we can do our best to destroy the boards and see what their limits are; DPS-2000BB supply is rated for about 165A but have been tested to operate around 200A, so the board is tentatively rated for 240A. That's the max for the screw terminals and current-measure circuit, at least; hopefully the heavy traces hold up without external bus wiring. If additional buswiring is required for reliable operation at stupidly-high currents, we'll add that to the initial prototype batch and work on revising the design. Once the prototype batch is tested satisfactorily we'll have a few available. Some will probably ship to a few potential large customers and we'll probably have a few for sale to folks that jump on it quick, but after the design is proven it'll be several weeks before we'd have the start of a large batch brought to market. Honestly I wouldn't look for a steady supply before sometime in June.


Design is definitely an iterative process, and we've been lucky so far to not have any catastrophic issues so stuff could be immediately brought to market after initial testing. As we learn more and refine designs more, and develop better tools and processes, stuff will be easier and faster to manufacture. The next D750 boards we'll apply what we learn from the DPS-2000BB development, and hopefully not need external bus wiring on those. All told, we should be able to cut the manufacturing time in half (compared to the original V0.4 boards) while still maintaining, if not increasing, overall quality and reliability.

Starting in a week or so we'll have another full-time guy around to help with manufacture, and we're also constantly working on new tools and methods to ease the process, so once we get to making the next batch of D750 and the first run of DPS-2000 boards, they ought to come off the line fairly quickly. We're also looking into improving the cable-making infrastructure, which if we can pull off what we want to do, we'll be able to have custom-length cable orders and additional standard lengths like 24, 30 and 36 inch in addition to the current 18".

[Lucky Cris]

Woot! Thanks for the update. Patiently awaiting for those DPS 2000 boards

[oner]

Woot! Thanks for the update. Patiently awaiting for those DPS 2000 boards

Agreed Cris, i'll definitely be anxiously waiting for a release of these!!
you can put me on board for (2) of the DPS 2000 BTC or $!

thanks for all the hard work @ sidehack & others. 

[Searing]

Woot! Thanks for the update. Patiently awaiting for those DPS 2000 boards

also considering....a couple of these at the guess of a 30 or 40 price

thanks for your efforts

Searing

[sidehack]

The DPS-2000BB boards will probably be priced more around $60-70. Compared to the D750 boards, the PCBs themselves will cost about 3 times as much (heavier copper traces and about twice the board area); the connector is larger and more expensive; there are 24 total screw terminals instead of 8; the integrated fan controller and other peripheral circuitry increase the SMD parts count three- or four-fold which increases assembly time (and therefore labor cost). Even with that though, adding a $30 PSU you'll be getting 12V power for under about $50 per kilowatt and that's not really possible to beat anywhere.

Once again I'll remind folks, we could scrimp on labor costs by having them assembled by discount slaves in overseas sweatshops, or we can continue to do what we're doing (and what we will never *not* do) and hire Americans to do the work, where we can oversee quality and pay people we actually know to do jobs we can actually train them ourselves to do. That might drive up the price a bit, but we think it's worth it. We wouldn't be able to warranty stuff basically forever if we had people we've never heard of on the other side of the world putting our stuff together. Reduced profit margins are better than reduced quality control accountability.


Also we got word today that the boardhouse had a machine down, which will delay our test board shipment by a day or two. We should still have them in hand within a week though, and start assembly and testing. It doesn't really affect our schedule much at all. But anyone needing power anytime in the next two months or so, let me remind you we do still have many of the 750W boards and they do parallel together. Three of those boards will push at least as much power as a DPS-2000BB, with actual redundancy.

[Scoobypup]

I actually will be needing power in the next 2 months (late june/early July at worst), but I unfortunately already have the dps 2000 psu's. Are you saying there is no chance of having the opportunity to purchase these boards before July?
To replace my psu's with the 750s would require me to buy 12 psu's and boards. At that cost, I may just resort to soldering and shorting pins to get my dps 2000 supplies working. I'd much prefer to utilize your expertly created boards, but it all comes down to timing I suppose.
It's unfortunate there isn't anything the community here can do to assist you with this project (unless there is...), but it appears that you are doing everything you can and are just stuck in that oh so familiar spot that many successful small businesses often find themselves in when they suddenly become "popular". 

[sidehack]

Hopefully we'll have boards available by the end of June. It's gonna depend on how quickly we can start getting some of the parts in, which will depend on factory lead times and how soon we order the parts, which will depend on when we have the money to do so, which is entirely external. But we'd like to be shipping in June.

One of the things we could have done to speed the process is get investors or sell preorders, but we don't really like either option. Especially not selling preorders before the product is fully designed - we won't let someone else pay for a product we don't actually have yet. Worst case, we might open up preorders with a guaranteed delivery time AFTER we have fully tested and working prototypes, basically in order to help raise enough money for bulk parts orders.

[peruana]

Do you guys have anything for HP Common Slot power supplies?

HP 800W POWER SUPPLY 399771-B21 403781-001 379123-001 380622-001 379124-001

[pmorici]

Do you guys have anything for HP Common Slot power supplies?

HP 800W POWER SUPPLY 399771-B21 403781-001 379123-001 380622-001 379124-001

I sell a board for that supply http://www.gigampz.com/ we are out of stock but expect more in next week.

[sidehack]

We were considering prototyping a board for that one but didn't feel like fighting him for marketshare since he already had something that worked. So, no. Buy his board instead, I guess.

[klondike_bar]

Do you guys have anything for HP Common Slot power supplies?

HP 800W POWER SUPPLY 399771-B21 403781-001 379123-001 380622-001 379124-001

I sell a board for that supply http://www.gigampz.com/ we are out of stock but expect more in next week.

I have two of these - nice build quality - simple but effective

[sidehack]

Oh yeah, prototype PCBs for the DPS-2000BB boards came in and we've assembled a board or two for initial testing. The dummy load we're building for stress-testing is functional but not functional enough to run the boards up to capacity yet so we haven't done full testing, but should be able to this weekend.