Bitcoin Forum
December 05, 2024, 01:19:52 PM *
News: Latest Bitcoin Core release: 28.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: BFL Single problems (electrical issue?)  (Read 1512 times)
mitty (OP)
Sr. Member
****
Offline Offline

Activity: 359
Merit: 250



View Profile
April 22, 2013, 05:11:18 AM
 #1

My mining rig consists of 37 BFL singles (FPGAs), and I seem to be having a lot of stability issues recently.

For the most part the rig is stable, but every once in a while one or two units start repeatedly giving me "invalid nonce - HW error" in bfgminer which decreases their hashrate based on accepted down to 400-500 MH/s (opposed to 800 MH/s) or sometimes lower.  Sometimes units do this and then die completely, despite firmware reflashes, requiring me to send the broken units back to BFL to be repaired.

I already tried reflashing the firmware (slower, faster, pretty much tried a bunch of them) on each problem unit, replaced USB cables with new ones, tried using a different computer as mining host, plugged them into different USB hubs, made sure temperatures are okay, and took units out of the cases to improve airflow.

Yet the "invalid nonce" problem keeps coming back :/

I feel like I've ruled out almost everything, but one thing that's really bothering me is that I think my apartment has an electrical grounding issue.  I'm running these units in the basement of my apartment.  I ran 4 new 120v 20a circuits to power the miners, have a UPS attached to each circuit, and have relatively balanced the BFLs across the circuits.  Each circuit has a GFCI receptacle to comply with electrical code regarding outlets in unfinished basements.  I'm pretty sure I wired the circuits correctly, but sometimes I get shocked when touch the case of the mining host computer or the ends of USB cables while standing barefoot on the cement basement floor.  I also noticed that I get a slight shock if I plug my laptop in to one of the mining circuits with its 2-prong (ungrounded) power adapter.

I checked and the voltage from the neutral to ground conductors on each circuit is minimal (less than 1v), leading me to believe that the circuits are wired correctly.  I did occasionally measure a voltage from my hand while standing barefoot on the basement floor to the neutral connector though, leading me to believe the house is improperly connected to physical ground.

The ground bus bar in the circuit breaker panel for my apartment is attached to the neutral of the service entrance, and to a copper wire which is attached to the incoming copper water pipe for the house.  I don't know if there's a grounding rod outside.

All the FPGAs and host computers are physically located on a plastic shelf which sits on the concrete basement floor.  Everything in the basement is wired to the new circuits which I installed, however the host computer is hooked up to the network via a cable I ran up to my room in my apartment upstairs.  The network switch upstairs is connected with a 2-prong wall wart but connected to a circuit which I think is properly wired.  

Other observations:
1) I see interference/static on the VGA monitor attached to the host computer.  I am using a 15' VGA cable so this could also be a factor though.
2) It recently rained and it seems like the nonce errors went away for a couple days, only to return today.  (a couple days after the rain).  Maybe the ground being wet improved the electrical ground resistance?

I know this is a strange problem, and may not even be the cause of my BFL issues.  These units seem somewhat poorly built and unstable in general, but I'm just trying to rule out any factors that could be my fault since BFL is probably going to very quickly get fed up with me asking them to fix my broken units.

Also does anyone know how I could improve the quality of the ground connection for the house?  I want to avoid having to call an electrician or complain to my landlords as my landlords have been very lenient with regards to letting me do pretty much whatever I want so I try to not bother them.

TL;DR I think my place has electrical grounding issues; seems like the actual ground is at a different potential than circuit ground/neutral.  Would this cause problems when running BFL singles?

Thank you in advance for any help/advice!
Protagonus
Full Member
***
Offline Offline

Activity: 238
Merit: 100



View Profile
April 22, 2013, 05:51:53 AM
 #2

Ye sounds like the ground is poor / too far away from your location. Possibly an electrical device in the apartment "shunting" some (small) bit of current to ground as well.  Since pretty much all chips nowadays are FET based; they are particularly sensitive to static.  Even with just a poor ground, static alone can cause your random problems via variations in ground moisture and Relative humidity of the air.  Here, too high and no static; too low and no floor conductivity.  Add to this; you have created a rather nice capacitor by placing a charge carrying surface (Mobo) on a dielectric (plastic shelf), with a conductive surface on the other side (floor @ X moisture).  Albeit a small value; this can still increase the likely hood of a charge accumulation without proper grounding.  

   If I were to make a suggestion;  get a solid copper ground rod driven into earth as nearby as possible. Either through a window or a spot in the floor if possible.  Either way, run lines from your earth ground directly to your GFCI.  You could go a step farther and run a line from earth ground to your PSU  case and mobo (screw mount hole).  You could use standard solid core or stranded to GFCI and a smaller gauge insulated to PSU / MOBO; if you added those as well.  Gauge isnt' as important really; as there's no real coulombs of electricity being transferred.  This will guarantee real ground isolation to your critical components regardless of existing ground.

  You can go even one step further, by placing a non-conductive porous material (gray foam) under the mobo; or even an anti static pad.  Anti-stat pad would have a ground clip but even a porous material like the gray foam will reduce the relative surface area of the dielectric to a minimum; as compared to current plastic.

I'm pretty sure I wired the circuits correctly, but sometimes I get shocked when touch the case of the mining host computer or the ends of USB cables while standing barefoot on the cement basement floor.  I also noticed that I get a slight shock if I plug my laptop in to one of the mining circuits with its 2-prong (ungrounded) power adapter.

A note, it takes about 3,000v or so; to cross an air gap and make a spark.   Just to give you an idea how much difference in potential between you and the host were "floating" at that time.  So say we have a power Mosfet (low RDS ON) in a closed state to ground at the moment one touches the case / board.  If that person is carrying a sufficient negative charge (typical for a person), this can be discharged through the FET instead of ground; thus blowing it.  Ground and poor grounds tend to represent or carry positive charges.

The ground bus bar in the circuit breaker panel for my apartment is attached to the neutral of the service entrance, and to a copper wire which is attached to the incoming copper water pipe for the house.  I don't know if there's a grounding rod outside.

Older appt?  At one point code would allow main grounding to water pipes.  However during this time, iron pipes were used for entry and thus not the best ground long-term.
Here the rust creates a slightly semi-conductive layer that again allows a small "floating" of voltage above ground.

hope this helps
Inaba
Legendary
*
Offline Offline

Activity: 1260
Merit: 1000



View Profile WWW
April 22, 2013, 06:06:25 AM
 #3

I don't know if this really addresses your problem directly, but when dealing with a lot of singles on one machine, I've found that occasionally one single will be *really* noisy on the USB bus for some reason.  Removing that unit from the USB chain solves issues with all the other units.  You can sometimes connect it to it's own USB chain on another port and things are fine or split your units into two rigs, putting the noisy unit on a chain with less units.  I had a devil of a time keeping over 40 singles running until I split them up and put the noisy units on a chain of 20 instead of all 40 on one chain, then the units were rock stable.


If you're searching these lines for a point, you've probably missed it.  There was never anything there in the first place.
mitty (OP)
Sr. Member
****
Offline Offline

Activity: 359
Merit: 250



View Profile
April 22, 2013, 03:45:58 PM
 #4

I don't know if this really addresses your problem directly, but when dealing with a lot of singles on one machine, I've found that occasionally one single will be *really* noisy on the USB bus for some reason.  Removing that unit from the USB chain solves issues with all the other units.  You can sometimes connect it to it's own USB chain on another port and things are fine or split your units into two rigs, putting the noisy unit on a chain with less units.  I had a devil of a time keeping over 40 singles running until I split them up and put the noisy units on a chain of 20 instead of all 40 on one chain, then the units were rock stable.


Thanks for the advice- I'll try isolating the problem unit.  Is there any way to measure how "noisy" a unit is?
Also this unit won't mine with the 832 MH/s firmware; I tried flashing 832 and it sits there at 0 KH/s.  I reflashed with 816 and it seems to usually work fine, with the exception of sometimes having periods of nonce error issues.  This isn't the first time this has happened; the past times it's happened I had to send the units back as they died completely shortly after.

The problem unit also sometimes dies and takes down all the other units on the hub.  Restarting bfgminer usually brings all the units back up after this happens. (I have a script that kills and restarts bfgminer if the average utility drops too low)

@Protagonus: Thanks for all the info.  I think I'm going to try installing a grounding rod; can it connect to a ground line on an existing circuit, or does it need to be wired directly into the circuit breaker panel?
Also with regard to me getting "shocked"; I don't mean this as a static electricity shock.  I mean I can feel a shock whenever I touch my laptop while it's plugged in, and a shock whenever I touch a USB cable end.  Nothing's jumping an air gap (as far as I can tell); it's a continuous shock that happens whenever I'm in contact with a seemingly grounded connector/my laptop case, which worries me.

And yes to older apartment; the house was built around 1900 according to public building records online. (may not be accurate, but it's still clearly very old)
The incoming water pipe looks like it's copper but I don't know anything about the water main or conductivity of the ground soil.  It looks like there was an old iron pipe that's no longer used but the grounding clamp is installed on the copper pipe.
Oldie
Newbie
*
Offline Offline

Activity: 52
Merit: 0



View Profile
April 22, 2013, 04:10:21 PM
 #5

Am I missing something with BFL?  I thought they haven't shipped anything (from what I've been reading), how can you have that many in a setup?

/confused
FatMagic
Full Member
***
Offline Offline

Activity: 224
Merit: 100



View Profile
April 22, 2013, 07:04:27 PM
 #6

Am I missing something with BFL?  I thought they haven't shipped anything (from what I've been reading), how can you have that many in a setup?

/confused

They are talking about BFL's previous FPGA's - not their current ASIC line.

Oldie
Newbie
*
Offline Offline

Activity: 52
Merit: 0



View Profile
April 22, 2013, 07:47:51 PM
 #7

They are talking about BFL's previous FPGA's - not their current ASIC line.

Ah, thanks, didn't know that.  sorry for the newbie questions.
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!