Not super comfortable with NDA's. I've been sharing my knowledge here as a matter of course, and I'd hate to not do that. Service to the community. However if people consider the name of a chip used on a board to be restricted information then so be it; I'll continue to go it alone.

Moving on, with the Qberty design the Pi power goes out when I plug the Pi into the Qberty board. That's odd. However if I lay the board down at an angle to the connector so just the first two pins connect it comes on. Which means something is odd in the first 10 pins on the Pi-Qberty board.

More working today. I fixed the bottom pins on that sacrificial board Searing, now I'm going to pull the top pins to see what's up if I can't rig up a quick header.

C

Never mind, I see what it is. It's a 24.576mhz crystal, reason being the BBB will output that signal out from the board itself, while the Pi can't do that. So KNC had to create an external frequency base off that crystal.

Interesting. Did you put the crystal in upside-down qberty?

C

Edit: Weirder. Something weird is going on the first couple of pins. Not sure why, but different from the other boards. Odd.

For the rest of the group, I have a Rev 2 bridgeboard that just came in. If you have an original bb that's blown I can probably fix that but you also need a new Rpi (Amazon has em) (send it and the old Rpi over), let me know.

C

Indeed. I have seen a few Rpis that basically short their +5 charge pump circuit. Since this is powered directly off the main power supply unit through the controller board without a fuse you wind up with a big solid dead short. The weak trace is under the bridgeboard, that's why it blows up in the first place. Fixing the trace isn't difficult but you need to replace the Rpi at the same time otherwise the new BB will blow open too.

I have a rev2 BB and some other things here for this weekend, going to be a lot of work tomorrow looking at some of this stuff. But if someone on the planet can tell me the BOM for this board or the Titan/Neptune I would appreciate it.

C

Can you tell us what the part was (digikey) and the nature of the fail (why would it kill a Pi?) so I can fix some of them for people?

C

Tell you what: If you post a whole iso dump of your sdcard I will take a look at it on my Rpi here and see if the logs make any sort of sense or if the programs can be kicked to do more loggging.

PM me a dropbox link or something with the file, I have a lot of Titans to work on but can put a look see in the work queue.

C

If they're having code problems they should just drop ship a single unit to Tarkin or something. The $600 or whatever cost is miniscule to the value of having more eyes on the problems.

Hm. Is the source controller computer a beaglebone black? I have a few of those over here and can fire it up to see what it does.

So how are the A4's going? If someone wants to send me one to test/troubleshoot/take apart let me know.

C

I use and like the Corsairs. 750 is the biggest I will use, if I am using bigger on a set of systems then IMO a short can cause a fireball before a shutdown. For normal units I'll go with Corsair 500's per each system.

Only Corsiar failure I have had was when the shed door blew open in a hurricane and the PS was heavily rained on. The chilli attached to it was also dead, but cleaning it up got it running again.

Neptunes have never done this. I have a bunch of boards and have tried to replicate this failure, no luck so far...

Also I'm going to be out in Vegas later this week, if anyone there wants to grab a beer let me know.

C

Interesting thought, I'll have to find some broken AMD cards somewhere. :-)

In the meantime I have found something interesting: Was looking at one of my sacrificial boards, pin 6 was dead shorted and I had drilled holes trying to clear the short. So I pulled the die to see where the short was on the chip.

The board was still shorted. Hm. Then I pulled the chip from the board where I pulled the dies out. Board still shorted.

I think I'm going to sit down and remove every component on my third board. If anyone has any shorted Titans they want me to destroy to figure this out let me know. (note: This is the pin 6 shorted issue, not the pin 4 shorted issue.)

C

I have seen pics of spoondles with burned connectors. I have fixed AsicMiner blades that people had shorted the fuses on (oops). I have pried connectors off BFL Monarchs that were pulling <250w per plug. And I have chipped off wreckage from both Neptunes and Titans that were run at 60mh/s (240w).

I have also worked on units that were dusty, dirty, filthy, and buried in manure. I have worked on units that were run in closed rooms in 120 degree aussie heat in the summer down there.

If and when the connectors implode on these things (if they wind up making it to market) I'd be happy to take a look and see what's left.

Some key things to remember:

0) Miner designs suck. Show me ones with UL listing for example, or even proper fusing.
1) It's not watts that determine the ampacity of wire, it's amps.
2) Amps go up as voltage goes down. This is very important when you find your 12v supply is only putting 10v at the miner plug.
3) Voltage drop increases as heat increases
4) Over time connectors get soft. Then one of the six pins starts to get a higher resistance. More current flows through the remaining ones, making things warmer.
5) Begin feedback loop.

Now commercial miners can run plugs with big wires going to 3,000 watt power supplies that will not sag if there is a heavy load on them. Plugs will hold up better. That's nice. But that same 3,000 watt supply will not sag when a cap fails in the miner and it turns the cap into a pretty plasma ball. Or mining chip or via, or whatever. That gets very funny very quickly.

Your mileage may vary. I don't argue with people who know everything, that seems to be pretty common on forums. I prefer objective reality and letting things play out.

Once again I wonder if the signal lines are isolated on these. That's always fun.

Fast summary: The damage to the plugs is cumulative as well as catastrophic. One of the odd things about the Titans is that they are one of the few miners that have had an operating profit life greater than a few months :-)

I'll do a bigger write-up later, but it's a combo of things that can make the plugs melt. Positive feedback loop.

Well I guess I should start quoting prices to fix the burned PCIe plugs that are going to come from this. .3btc per plug burned plus shipping should do it as long as they didn't put an insane amount of copper into the board underneath.

And if you're thinking just chip off the burned plastic and get back to mining, don't. Either you will plug in the miner wrong by mistake; blowing out everything on every line or the damaged and pitted pins will lift a ground resulting in complete hilarity or you will overheat a pin burning out the internal vias to the power planes, causing all sorts of havoc.

Now for the fun question: Did they put optoisolators on the signal lines to prevent the inevitable lifted grounds from causing ground to rip though either the hashing chips or the level converters? Time will tell....

C

Why does this sound familiar?

Yeah, I think it's sad that home mining is going away. The root of the problem though is simple: Mining is a perfect market, and as such it's not possible to make profits. Home people have it worse, mining hardware by nature is buggy and experimental and really crap in terms of quality. It rewards people who are willing to take shortcuts (powering a bunch of miners off a big PS without fusing which equals fire but better efficiency), and it rewards companies who put out the cheapest fastest developed things possible (since checking for errors takes time and time means you lose the window to market to someone who didn't bother to check).

KNC did this with the lack of optoisolators on the signal inputs. Had they done this, 90% of the problems would not have happened. But you can't blame them; spending that time to put in that solution would have put them behind someone else who didn't give a fuck.

It's a queen's race right to the bottom. Drat.

I guess, and if someone really wants to move it, go for it. This started as a bitcoin thread with the Neptunes and Titans are basically Neptunes that make money and have different chips.

What should I look into fixing next, if anything? To be honest, bitcoin mining gear depreciates so quickly that by the time you ship it, it's obsolete :-)

No problem, just busy on a couple of other projects. And seriously hating trying to do diagnostics on neptunes (because they're kinda like titans without the value)  using my Tektronix scope. It is a bitch on wheels trying to probe SOT type chips, I'm going to have to micro-solder wire wrap to the pins to get test points.

Bitch on wheels.