On the board whose fan speed appears to have stopped working, make sure it's set to internal instead of external control. If it's set to external, the knob signal will be blocked from the PSU and it won't do a thing.
As for parallel, I have tested three supplies tied together working flawlessly to power five overclocked Ants. Within the next few days I'm going to test up to at least six in parallel and see how they behave. I'll be using the new V0.5 boards so I have accurate current measurements from each board.
|
|
|
We had the revisions made for the V0.5 board the night we received the V0.4 batch. Put it together, ran it through the ringer and made a list of fixes. I just made the first V0.5 fully-functional board last night and we have a list of stuff to change for the V0.6 as well. Gotta keep making things better.
As for the workers, well they're gonna get a full retraining this week on surface-mount soldering and general quality standards. One of the guys is already putting out pretty good work. He's my least-experienced minion, with the least free time, and is putting up more productivity and the highest quality of work.
|
|
|
I just tested a modification on the next batch of boards that should give touchy but reliable fan control to a wider variety of supplies. Just tested it on N750P-S0, N750P-S1, Z750P-00 Rev A00, A01 and A02. They're boards we should start shipping probably Wednesday. ( https://bitcointalk.org/index.php?topic=503423.msg5754835#msg5754835) Yep, it's been addressed and tested. As for running on no fan, I don't recommend it. I tend to run them on a low setting still above the bottom-end of the adjustment. The only supply I've ever had fail on me was partly disassembled, and getting almost no airflow over the components at the DC end (which are used to direct fan exposure); after running probably 85-90% load (two custom-overclocked Cubes) for a little short of a week it powered down for good. Run super-low-fan if you like, and it might work just fine forever, but officially I don't recommend it. At full load you need to be able to dissipate between 70 and 80W of waste heat.
|
|
|
Oh yeah and they'll have a better mating connector to plug into the PSU. That's one of the things I wasn't really satisfied with on the V0.4, but we got some better parts and a better board to fit them. And the terminal blocks and header pins are labeled better on this version so it'll be easier to know what you're doing. The toggle switch mounts better so it'll be more solid on the board, and hopefully all the underside work will be cleaner. All in all, I'm pretty pleased with the V0.5 boards we got now.
Oh yeah and the current meter should be accurate now. It's calculated to 1.8V on the CUR pin corresponds to a 60A load, with a 5% tolerance on the measurement.
|
|
|
I just tested a modification on the next batch of boards that should give touchy but reliable fan control to a wider variety of supplies. Just tested it on N750P-S0, N750P-S1, Z750P-00 Rev A00, A01 and A02. They're boards we should start shipping probably Wednesday.
|
|
|
I'm actually gonna have to wire up a new 20- or 30-A 220V outlet so I can stress-test the DPS-2000BB boards when we get them in and ready. It'll also come in handy for future mining and server stuff. Hopefully I'll have prototype DPS-2000BB boards in hand inside of two weeks. Unfortunately faster also means stupidly expensive.
|
|
|
I do have some of the other revision supplies to play around with. Also the remaining parts we'll need for the V0.5 boards should arrive about Tuesday and we can start manufacture, so we'll probably be shipping them in about a week. I'll have to retrain the minions, which hopefully will result in (after training) faster turnaround and a higher quality of work so's we can get you guys your stuff quicker.
|
|
|
I'll have to compromise on the next board since the PCBs are already delivered, change parts around a bit such that you'll have full control on the ones you currently have no control over, and have full but not as precise control over the ones that currently work fine. Probably the V0.6 boards will address the different range standards with a switch setting, but for now we'll have to make do.
|
|
|
I think bobsag said they ran a couple 400MHz Ants off a single board for a week and some without issue. I haven't tried it but apparently it works.
|
|
|
I'm finalizing the PCB design for a DPS-2000BB interface board this week, and hopefully next week we'll be stress-testing some quick-turn boards to make sure they're good for 200A. Should have a DPS-800GBA board as well, for folks that might be retiring their Blade backplanes and want to run 3 Ants off that power instead.
|
|
|
Snazzy. Hopefully we'll be stress-testing the first DPS-2000BB PCB prototypes in a little over a week. It'll have a built-in controller for two 4-wire fans with full internal and external speed control, as well as the standard 5V/3.3V, current share and external-on lines. Building the test rig is gonna be fun. 12V 200A is a lot of heat to dissipate. It doesnt get any better than this. Well... I'm sorta banking on that it does... but for the basics, that's a pretty slick setup.
|
|
|
I wasn't sure if I had explicitly explained it here or not, so if I haven't then you have no need to apologize.
|
|
|
First thing to try - make sure it's set for internal instead of external fan speed control. If it is set for external, set it to internal and try again. If it still doesn't work, set it to external and ground the FAN pin. If it still doesn't work, I have no idea.
If it doesn't work on internal but works on grounded external, short around the 393 resistor on the backside of the board and try the knob again (on internal). If that works, either permanently short around that resistor or tie a wire from the rightmost pin on the knob to ground.
The other option if it works on external, is rig up an external pot between the 3.3V and GND pins on the header, tap on FAN,and adjust from there.
For the record, this is not a problem with the board. It's a change in internal design spec on a different variant of the power supply than I had to test with. The half-dozen supplies I used all worked, but there exist other variants that, for some reason, use a different voltage standard for fan control. It's not a problem if you have a closed-loop active controller, but that's not what we have.
And since this is the third time I've answered this question today, and at least the eighth time I've answered this question in the last week, I'm gonna start directing people here if they ever mention the word "fan speed" to me in any communications.
|
|
|
Unless you ever need to maintain it.
A semi-permanent solution I'd recommend is cut a piece of plastic or something and mount it using the screw holes on the board. Won't come off on its own but easy to remove if you need to.
|
|
|
Yes, if the bottom side hits metal it could risk shorting. I'd recommend putting a rubber mat, tupperware lid, or junkmail underneath the board just in case.
|
|
|
10 Blades, 5 two-blade (standard) AntMiner S1 units.
|
|
|
I know when the servers fire up they run at full fan everything ever for a few seconds, until stuff is initialized and it throttles them all down to manageable levels. Two PSUs and a couple other 2U fans and those servers are effing LOUD.
Also, a consideration I'd like opinions on. The DPS-2000BB does not have internal fans, so external fans need to be provided. I'm going to put a speed controller with fan headers on the board, probably for a 4-wire fan. Does anyone have a compelling reason why I should design for a 2/3-wire fan instead?
|
|
|
I just took one off my desk that's been running in my room since probably November. It does definitely seem quieter in here, but not enough that I really care. Maybe that's what happens when you work around loud fans every day for three years though. When the fans are turned down they're not too bad. By my standards. I'd rather have stuff that works really well than is comfortable.
|
|
|
I don't have any DPS-800 boards for sale, we're still prototyping. All we have right now are the D750 boards. Load-balancing is pretty easy once you understand what's involved. Load-balancing and master on/off can be done with two power wires and two signal wires. https://bitcointalk.org/index.php?topic=503423.0
|
|
|
|