Bitcoin Forum

Morning all!

SOLUTION FOUND. REPLACE THE FT232 CHIP AND YOUR JALLY WILL LIVE AGAIN. GET SOMEONE WHO KNOWS HOW TO DO REWORK LIKE XBOXES TO DO THIS FOR YOU. THE PART IS $3.00.

I've been hacking away at my BFL Jalapeno and have been noticing that a number of people have been saying their Jalapenos (jallies) have been failing with either a power supply failure, or a failure to connect to their desktop computer with USB errors. Or both. Or they get a new supply and now their jally isn't talking.

This seems to be happening in the EU, England, and Hong Kong. After reading the reports and such I think I know what the problem might be, and will be updating this thread as it develops. I'll keep this post for summary advice, and put the details below.

NOTE: I AM NOT RESPONSIBLE FOR ANYTHING YOU DO, DON'T DO, OR THINK! These are only the observations of a random Internet user, please don't take my advice as legal counsel, gospel, or whatever.

Anyway, what I would do if I was in another country with 240 volt mains is this:

1) Get another power supply on your Jally. Get one that is UL/CSA/Whatever your country considers a standard for grounding and tolerances. This would be a good idea IMO.

2) If you can get a true Isolation transformer, plug the jally into that depending on local codes and laws.

2) Put a USB Opto-isolator between your jally and your computer. Note I have not tried this yet, as I am in the US and this is not a problem, but I *think* this will prevent the USB from blowing up. Something like this:

http://www.amazon.com/Multi-purpose-USB-Isolator-Built-in-DC-DC/dp/B00899UVGW/ref=sr_1_1?ie=UTF8&qid=1386683733&sr=8-1&keywords=usb+isolator

Back when I was working on high voltage DC electric car stuff, you would always use optoisolators on the RS232 segments to prevent a ground loop from causing problems.

3) If you have GFCIs (Ground Fault Circuit interruptors), plug your stuff into one of those. I don't know power in Europe, or if thes exist.

Regardless, if you have a Jally outside the US and want to comment, please feel free to do so. I'll update this thread with some thoughts.

Thanks!

This is a ground fault.

A GFCI or related unit will NOT protect your device.

A USB isolator might. It does prevent a ground fault from happening.

Someone confirmed that a GFCI trip happened, and now his device will not recognize.

If your jally comes up with the front LED flashing fast, the power supply can't provide enough current to pass the diagnostics. Unit isn't bad, the problem is the stupid supply is bad.

Likewise I'll bet the reason that singles are coming up without chips is due to power problems. Get a real 500 watt power supply.

These have been confirmed by experiments here at danger labs. Flat facts.

You don't need to do any of that crap.  Just swap the supplied Whooflugdung PSU with something better, be it LiteOn or whatever, or run it off an ATX PSU like I do.

Update: I've been getting reports that the power supplies are the US supplies with US cords, and people are plugging them into local power plugs with adapters.

Note: I think that this is not a good idea. Your local power supplies probably have 3 pin grounds, they are there for a reason. If *I* were there, I would use an adapter that provided full and proper isolation.

More to follow. Does anyone care by the way?

That was option 1. However I don't think most people understand what is going on. And it's not a "better" supply, it's one certified to run properly in your country.

C

No, because you're making a mountain out of an ant hill.  Please stop spreading FUD.

Really? If this is not an issue, I'll pull it.

Actually, you can help: If you know how to do it, and can check voltages between each of your power legs to "ground", I would appreciate it. Do you see 120/120 on each leg, or is it 240/0?

This image is what is bothering me:

https://images.weserv.nl/?url=fbcdn-sphotos-h-a.akamaihd.net/hphotos-ak-prn2/1472782_10153499628905184_918095915_n.jpg&fnr

I have at least one EU person who is seeing 120 volts on the ring of his jally power supply to ground. Either his picture is lying, or 120 volts is on the ground plane. If that were to happen, and the ground plane on a desktop attached via USB were to be grounded, then 120 volts would merrily go from the jally through the USB to ground. I think this would be tough on the Jally's USB chips.

I think this is why EU laptop power supplies have 3 wires, hot and explicit ground. But I have heard that the power supplies shipped are US supplies with two wires.

C

I wouldn't call the issue FUD, there have been a large number of reported unit failures, and in 9 out of 10 cases I suspect the blame can be pointed squarely at the power supplies. This is pretty simple really, the BFL units are purportedly tested before being sent out, so if they had any issues they would fail in testing rather than a few hours after customer unboxing. The power supplies however are not part of the testing. Look at the test rigs BFL showed in their video, they have a much bigger common PSU running everything instead.

My guess is the PSUs are not tested and are poor quality chinese products made as cheaply as possible.

Agreed. I just checked some threads at BFL, and there are a number of people with the same problem. All seem to be in the Europlug zones.

Now BFL is sending out RMAs and such, they're fine there and meeting their terms of support which is good. But if there is something that can prevent this from happening so you're not down for weeks, I would think some BFL buyers might be interested.

C

It's not the problem you think it is.  There might well be 120V there, but there's probably insignificant current.  If there were, the RCD or RCBO in the house's electrical system would trip. 

If you read the thread that photo came from, this is exactly what happened. The PSU killed the circuit and it needed to be reset to bring all the various household appliances and things online. The problem was traced to the PSU. This measurement was taken when the PSU was powered on, plugged up to the Jalapeno but not to the host PC. Connecting it to the host PC caused the circuit to trip, and with a potential like that between the ground, I can see why!

So the PSU had a fault.  ANY PSU can have a fault.  I've had brand new, out-of-the-box Apple laptop chargers that have died, same with Dell, and HP. 

But, my sentiment still stands, just because there's mains potential between the output ground and mains ground, doesn't mean there's a fault, or that there is a danger.

Indeed. The problem isn't faulting, shit happens. The problem is the impact; does the device fail safe or fail destructive.

Fail safe would be if a fault shorts to a ground at the PSU, blowing a fuse (in the PSU) or a main breaker. If you have a grounded device and your wiring meets NEC (or whatever you use in your country that is not corrupt) your PSU dies and you get another one.

Fail destructive would be if a fault shorts, there is no ground, and the case goes hot (deadly), an exposed connector goes hot (painful), or the fault goes to ground through another device (destructive). In the US, BFL supplies can't fail deadly. Those same power supplies on a 240 volt hot hot ground system would fail destructive.

This thread is for information, not judgement. Given that bitcoin mining is on bitcoin time, and difficulty increases, it might be better to spend $40 on a power supply and not be down for a month than to risk failure and be down for a month on an RMA.

As I like to say in Bitcoin world, do the math :-)

C

From https://forums.butterflylabs.com/jalapeno-single-sc-support/5989-safety-issue-jalapeno-causing-power-trips-sparking-usb-port.html
I don't have one of the failed ones here, but I'm not so sure they're failing in a safe way. A power supply that fails in such a way that gets a strong shock through their arm by touching it is extremely dangerous.  BFL needs to stop shipping those supplies, the reports of Jalapenos tripping home circuit breakers is disturbingly numerous.

Hopefully (and some bitcoin later) I will be having one come in here for review. I think it is fail-destructive, which is not good. But I think any power supply could do this, the problem is the ground loop back to the computer which is grounded equals the usual boom.

C

Generally, a device should either be safety earthed o the chassis can't be energized, or be double insulated. In this case, it would be a double insulated device. They are designed so that no single fault can cause the mains voltage to be exposed. A good power supply should not fail this way even if it does die, it should just stop working. There is obviously something very wrong with these ones as they appear to be able to provide significant mains current at the output, and still function otherwise normally. It will be interesting to see what you find, do you have much experience with mains powered SMPS designs?

Bit. My expertise is in PFC charging systems and the good old fashioned "Use the motor as a big coil" charging systems on electric cars with 50-80kw power drive systems. Thus I look at the 1 volt supply and think "If I replaced that with 500amp IGBTs, I would be in shape" then realize that would be insane.

But these problems are not unusual. When charging on 240 US you have to watch for ground faults that can cause 120 to appear in odd places. My guess is they have the neutral connected to 12 volt negative with a capacitor to improve the power factor. The PF on these supplies sucks rocks regardless; I can see that on my P3. That's why they get hot, that whole reactive power thing.

I just want to fix this so people will not send in their jallies for repair which will mean my jally will ship quicker.

C

Have Cablez make you PCI-E power plugs for your Jally.

Problem solved, and excellent service.

That way, you don't have to deal with cheap power bricks of any kind.

Just did, takes a 16 gauge wire to really support the jally. Piggy power draw but mine has 5 chips.

Talking to two other people same problem. I hope I can get one of these to try fixing it.

C

Based on this post over at BFL I am virtually *positive* this is a ground fault. See:

https://forums.butterflylabs.com/jalapeno-single-sc-support/6812-jalapeno-bf0005g-trips-my-main-fuse-boards-residual-current-detectors-rcd.html

Note also that an isolator prevents the trip, I think it would save the jally.

C
 

I've blown 3 out of 4 of those power adapters, I've converted all my devices to just power from my ATX power supply.  One of the power adapters was so bad, any time I would plug it up, I would blow a fuse at my circuit breaker.  I don't trust those things at all.

https://i.imgur.com/GjDu2oC.jpg

Nice picture; that solution should work for a normal jally.

C

I'm trying to do what you're doing as well by adding more chips to the device ( eventually 6 ).  I made my own adapters using 16 gauge wire, is there any reason you'd think that I wouldnt be able to power 4 chips per device?

The ATX power supply is 1200 watts, and the cables are 16 gauge.  I will admit to have blowing a few breakers as well, but I cant see any reason it wouldnt be able to support 6 per device on a 16G cable.

The only issue is voltage drop; I noticed the wire gauge on the jally cable was 16 or so, and the wires from my cheap-o ATX were 18. Running it off 18g wire at 100 watt draw (it's hungry) caused a 1.5+ volt drop, and would shut down the PS.

Tripling up the wires and shortening them dropped the voltage drop to .5 volts, which is much better. Running happily, but my noctura fan runs a bit slower, resulting in the bottom heat sink being back on the unit. Still looks good though.

C

Most modern power supplies have a single rail ( or multiple ) that you can easily upgrade.  I changed mine out to 14G because I was powering 6 asicminer blades and the cables were getting HOT.

Well I got the broken jally in the mail today and it is *exceptionally* interesting to see what's going on here.

The power supply was blown of course, no output, output caps exploded. However what is interesting is that although the case was factory sealed, after opening it I can see that the power wires to the jally were either soldered by a junkie, or were done by someone who hit the output caps as well. Lots of solder, crappy job, pics to follow. No wonder it blew.

More interesting: Plug it into a big power supply, it fires up, lights the two chips, fails to talk. No biggie, waiting on my ftdi chips from mouser. Ground plane is intact, not that. However VERY INTERESTING THING I HAVE SEEN BECAUSE I HAVE COOL POWER SUPPLIES:

My power supplies are regulated output, 2amps max up to 52 volts. I set it for 13v, 50ma to start with. Hooked up, voltage drops to 3 volts, power LED barely comes on. Crank the current, at 1.5a it starts to come alive, the jally starts testing, and the voltage drops sharply from 12.9 volts down to 5. Then the jally fast flashes the front led.

This is something people see. In this case, the power supply can't provide enough surge power. So I ran the current up to 2.2 amps which is max. Tried it out.

Voltage drops to 8 volts unit flashes. So what's happening here is interesting: The jally starts up, pulls enough power to start, the pulls the power supply *hard* to bring the chips online. If there isn't enough current it will fast flash fail.

Which means if you get that error on a Jalapeno, Single, or Little Single, then the problem is the power supply is too weak. Don't return the whole thing, get a man's power supply and go.

Very interesting confirmation here. Edit: This is confirmed by hooking same jally up to my 500 watt corsair which drops from 12.4 volts to 12.2. So what I have written is a flat fact.

C

Update: Reflowing the FT232 chip did not help. To verify I heated each quarter with a soldering iron, then verified no shorts and good connections. Same thing, did not work. This is not a reflow problem.

Back to drawing board. Will swap out 232 chip this week.

Update: Took the old 232 chip off the board. You cannot do this with a soldering iron, the back of the chip is soldered to the board via a ground plane. Think air tools and temps of 450c+ for over a minute. That's a massive amount of heat.

C

Yes I floated the chip off like putting the ASIC on. Pre the area with some flux and heat again to flow smooth. Put new chip on, flow it onto the board. while its still hot put a little more flux and solder the leads on good. Clean up as necessary. (A short wouldn't make ya happy right?)

I think this was the hardest chip to put on yet. My forth jallie when given enough power by a 1500 Watt PSU lit the power LED for a sec and is now dead again. Hmm back to square one with out a hole in the chip this time. I want to know if anyone successfully reballed a chip yet. I don't have a problem with stealing the chips then.

Haven't reballed, but I have taken the solder off two chips, and will try mounting them to the next jally. I think you can do it without reballing, my problem is cz main jally is not going to take more chips. I think in your case one of the chips is shorted on the 1v supply. Fix it and mine like hell :-)

C

Ok, I've thought about this and the chip would sit at a completely different height look at your chips and the ones BFL put on. You can see with a flashlight and bare eye underneath. Theres a small amount of space created by the solder balls. If they say perfectly flat with solder those little balls would cover a lot more area. W/o the balls it will prolly cause heat issues from 2 things the heat not being dissipated into the sink right unless you shim, and the bad contacts on the BGA can cause numerous problems like the connection occasionally bridging or burning out. I don't even know what size balls these are or if they make a template for the masses to easily reball with enough time a template may not be needed.

As for the dead jallie I think you're right about it being a short. I don't know the places to look, I've literally stared at it for ever. It's a second rev board with only one JTAG so it may not lay it out as simple. I'm not following traces on the back of the board though. If I got a nice pic and started labeling board parts would you want to fill in where you can?   

Well, let's think: What could cause a short? Your chips!

If you don't want to take them off, then get a light, put it on one side of one of your chips, and look through with a loupe or glass on the other side. See if all the balls are perfect? Then do the other axis.

Try removing your chips. Yes, you're gonna reball. Let's go halfsies on a reballing stensil and table, screw it.

Myself, I have a vial of balls from ebay, since there are only so many of the things and I have watch tools I am going to try putting them all on and see if I can make it go. Line them up, then low low low air and heat to see if I can melt them. If it doesn't work I'll get a stencil.

As for the first paragraph, eh I work with heat pads which can make up the height loss. I'll try one on the dead UK board once I bring it back online.

C

Ok I've been looking around and can find no info what so ever on the chips ball size. What are you using? I'm fine with spending a few bucks but can't find any info. If I'm going to be putting on 144 balls on every chip I want to do it right the first time.

Um I forgot. However I put a new 232 chip on, nothing. Interesting. However a question: Do you see ground on any of the 4 USB pins?

Oh and 4 chip jallies with AL heat sinks? HOT AS HELL! Put sinks on the bottom like now.

C

I just checked on the dead one w/o chips. Using the usb housing as ground none of them were grounded. Do you want to know about the living?

If you can, please.

C

Jesus fucking Christ. You know, I think given enough time I could fix a rainy day.

Summary: This burned out jally with no USB response is fully operational. What happens is the USB FTDI232 chip burns out, and you have to replace it. For someone who has done Xbox repair this should not be hard. For someone like me who has done BGA and SMD work but not these stupid QQ chips it's a screaming pain-learning curve.

Removed the old chip with air tools at 450c with a 350F preheater under the board. Came off no problem. Put new chip on with same temps (2 mins at 450c) like a BGA cpu. Burned the fucking chip out.

Pulled it, put another chip on, 275c for 60 seconds with a 350f preheater (I buy 3 for shit like this). Chip goes on, comes up as a BITFORCE, all "J's" when I try to run diagnostics. Um, I think I need to reflow. Added solder balls, melted with the iron, cursed a lot, hit it with 300c heat for 60 seconds around the sides of the chip, and...

It COMES UP! It's a 6gh jally. It sucks, but it does run.

So I fire up the BFGMiner and it immediately screams that the chip is at 130C. Fuck again. Check the diags, S2 is at 130, S1 is at 30c. Um... Pull the heat sink. Note this one had a loose heat sink from the factory and it was banging around. Found that S2 was ripped off the board by aformentioned sink, and the pads were gone too.

Time for a lobotomy. So I went into the code, changed S2's AD channel to be S1 so the code will always read the working S1, put in a comment saying I know this sucks, but I had no choice due to bad assembly, compiled it, loaded it, attached the fan and...

It's hashing away at 6gh now. So it looks like a blown European Jally can be fixed with a $3.00 part, and a person who knows how to sweat chips on and off the board.

Go now, everyone. Fix your jallies, hash, and bring the difficulty up. I have completed this task, and if it wasn't 4am I'd say it's Miller time.

Fuck it, it's always Miller time. Off to get a cold one then to bed.

Update. Sunday night, still running, no problems. Glad this works.

C

After a quick read thorough, my 2 cents worth.

Am sure alot of these units will have been pluged in using a US/Euro adapter.

The PSU I received is auto switching so voltage is not a problem.

I have seen many of these adapters that do NOT carry an earth/ground pin. In this case the PSU may technically fail safe but the adapter is preventing the fault path to achieve this, raising the potential between the chasis ( and USB GND) to the ground. OUCH. So the fault path becomes dependent on the USB connection to the PC, be careful plugging that baby in  :o

Easy solution, cut the plug off and put a local one on, seek advice on connections

Brown = Live/Active
Blue = Neutral
Green/Yellow = Earth/Ground

or

red 2 red , black 2 black and blue to bits   :D

Thank you very much for your report, lightfoot.
The same thing has happened to my friend's Jala. Power adapter failure, and the unit not recognized thereafter. Booting normally, LED flashing first, than steady, two LEDs for two chips... So I believe the PSU has taken the USB chip with it, as you have found and verified.
It's not quite simple to obtain a new one though. There are several types & packages of FT232, but the HQ one is not frequent at all...
Mouser wants 40 euros (!) for p&p (or I've failed to find a better offer), maybe Farnell will be more acceptable, its shipping is "only" 150% of the price of the component itself... :-))) I've left this on the owner of the cube. I myself don't have one unfortunately... :)

BTW, what's Miller? :) A sort of beer, I guess? ;-)

I think the problem is the power supplies have a capacitor going from one of the two AC pins to the DC neutral. This can be very helpful for increasing the power factor, but if the capacitor were to short then it would conduct one of the 240 volt legs to the negative line on the jally.

This 120 volts would not be a problem except for the fact that it goes to the ground of the USB line. Which then goes to your computer. If your computer's USB jacket is grounded, then 120 volts goes from wall through broken capacitor through jally through USB cable to computer to frame ground back to the utility. Boom.

Speaking of which, one should be able to replace this chip with an LQFP 48 pin one which would be a hell of a lot easier to solder. Recommended for followup.

C

How about the HF? The things cost 5 bucks here in the US. What's with the high euro price?

C

Suppose the capacitor goes from one AC pin to the DC "ground", as you say.
Then, in Europe, you won't get 120 V there. The 230 or 240 V lines aren't symmetric to the ground. When you're lucky, the capacitor is between AC neutral and DC neutral. When unlucky, you get full 240 V there.

Another possibility might be there's a "bridge" in the PSU, halving the input voltage, and its center is connected to the output neutral... This way it could produce 120 V in 240 V environment.

The LQFP package has a little bit larger outer dimension (ends of the pins to the opposite ends). So it won't fit on the PCB pattern exactly. The pins will sit on the PCB pads, but will exceed them a bit, so they can't be soldered standard way as on their regular pads.

There is a fair bit of extra pad, might fit. Worth a try...

The price of the coponent itself is OK here too, but the p&p is sometimes unreasonable.. :-)
Maybe we'll try the HF one, if it'll be more easy to get. Hope we can try today.

Bonus style points if it doesn't, but you bend the leads over like a PLCC.

I bought two spare PSUs for my Jallys on Amazon for $8.00 each.  They work great.  BFL also sent me a replacement for the first one that died.  I just figured it was worth $16 to keep them up and running while waiting.

Makes sense. With these three new jallies, I ordered the 3 wire PCIx cable so I wouldn't have to use the power supplies. Guess what didn't come :-)

Oh well, so I get to cut off the wires and hook them into my corsair. Man that thing is under a serious load right now, especially with 20gh per jally.

C