Block Erupter USB - Overclocking/ hacking ?

[mjgraham]

My latest incarnation, decided to build a circuit board, plus as USB hub controller plus an aftermarket dc-dc , working the speed back up, I have found the problem that is causing my high speed issues seems to be ground loops, I guess it is to be expected with so many random parts in play. seems to be on the USB side I guess the long cable to the raspi that has a ground too. looks neat but doesn't work as well as I had hoped. I actually had all these parts already for many projects in the past. Have a single clock with a fanout buffer to the others. Plus all the work involved for such little gain really it changed it from ROI (just for the BE) in 100 days to like 70 ad the price I paid for them, buy then may not even do that. I think the best plan is to use the 16 Mhz and the resistor change far cheaper and effective. If I could get them to go to 28 Mhz would be lie having 7 for the price of 3 might be OK but I don't think I'll make it. As a friend of mine said "your pushing the limits of man and machine"

[1714989492]

1714989492

[1714989492]

1714989492

[1714989492]

1714989492

[1714989492]

1714989492

[cableiso]

Switched to bfgminer 3.1.4; will post results soon. Miner feels very hot.

UPDATE:
ICA 0:       | 587.6/441.8/382.4Mh/s | A:127 R:0+0(none) HW:34/ 21%

[/quote]


Yes, 21% HW errors is too much.  You should try to resolder R1 to make sure you're not overvoltaging the device, and improve the heatsinking. 

[cableiso]

That is freaking fantastic!  Is that a Vicor DC-DC module there? 

Anyway, sweet rig - and taking off the inductor is a great way to feed the LV right into the device.  Brilliant!

My latest incarnation, decided to build a circuit board, plus as USB hub controller plus an aftermarket dc-dc , working the speed back up, I have found the problem that is causing my high speed issues seems to be ground loops, I guess it is to be expected with so many random parts in play. seems to be on the USB side I guess the long cable to the raspi that has a ground too. looks neat but doesn't work as well as I had hoped. I actually had all these parts already for many projects in the past. Have a single clock with a fanout buffer to the others. Plus all the work involved for such little gain really it changed it from ROI (just for the BE) in 100 days to like 70 ad the price I paid for them, buy then may not even do that. I think the best plan is to use the 16 Mhz and the resistor change far cheaper and effective. If I could get them to go to 28 Mhz would be lie having 7 for the price of 3 might be OK but I don't think I'll make it. As a friend of mine said "your pushing the limits of man and machine"

[Dunkelheit667]

...I think someone said 3.1.1 was the last version that they worked in, I did not test it,...

Just to clarify. Guess it was my guess and I was wrong. --icarus-timing short and --icarus-timing long still working fine with latest cgminer versions. Mea culpa. But back to topic.

[Scorpia]

...I think someone said 3.1.1 was the last version that they worked in, I did not test it,...

Just to clarify. Guess it was my guess and I was wrong. --icarus-timing short and --icarus-timing long still working fine with latest cgminer versions. Mea culpa. But back to topic.

yes with the latest versions 3.4.1+ the timing option does work and an extra option has been added. i use --icarus-timing short=90 which gives me 447 on all the 16Mhz units.

it also stops the timing issues is the PC has issues which running short.

[Scorpia]

R1 = 1.2K
Q1 = 16Mhz

No longer works in raspi + DUB-H7 setup, cgminer sees the device, not able to run it properly, I suspect lack of power (5 other devices are running in the hub, this one plugged into "high current" port)
On standalone PC cgminer 3.4.2 sees and runs, not able to get over 350Mh/s with icarus-timing short

Bitminter on Ubuntu reports response time of 6ms and 447.9Mh/s hash rate, average is 447.0Mh/s.

Is it what everybody got or I've done something wrong?

I don't think the options in the newer version of cgminer work anymore, I think someone said 3.1.1 was the last version that they worked in, I did not test it, but all versions of bfgminer I have used works fine with the -S all option

Switched to bfgminer 3.1.4; will post results soon. Miner feels very hot.

UPDATE:
ICA 0:       | 587.6/441.8/382.4Mh/s | A:127 R:0+0(none) HW:34/ 21%

im would think that a 1.2K resistor is to small and yu are undervolting the chip for that speed. i would suggest a 1.35K ohm (ish).

also yes make sure you are cooling them with a fan at least.

but at 16Mhz and upto 1.3v i have ran my sticks for days with little errors using a desk fan for cooling. cgminer 3.4.1+ using the --icarus-timing short=90 option

[cryptomining]

My latest incarnation, decided to build a circuit board, plus as USB hub controller plus an aftermarket dc-dc , working the speed back up, I have found the problem that is causing my high speed issues seems to be ground loops, I guess it is to be expected with so many random parts in play. seems to be on the USB side I guess the long cable to the raspi that has a ground too. looks neat but doesn't work as well as I had hoped. I actually had all these parts already for many projects in the past. Have a single clock with a fanout buffer to the others. Plus all the work involved for such little gain really it changed it from ROI (just for the BE) in 100 days to like 70 ad the price I paid for them, buy then may not even do that. I think the best plan is to use the 16 Mhz and the resistor change far cheaper and effective. If I could get them to go to 28 Mhz would be lie having 7 for the price of 3 might be OK but I don't think I'll make it. As a friend of mine said "your pushing the limits of man and machine"

Awesome rig you got there, thanks for the cool pics and the updates.

 

[Trongersoll]

My latest incarnation, decided to build a circuit board, plus as USB hub controller plus an aftermarket dc-dc , working the speed back up, I have found the problem that is causing my high speed issues seems to be ground loops, I guess it is to be expected with so many random parts in play. seems to be on the USB side I guess the long cable to the raspi that has a ground too. looks neat but doesn't work as well as I had hoped. I actually had all these parts already for many projects in the past. Have a single clock with a fanout buffer to the others. Plus all the work involved for such little gain really it changed it from ROI (just for the BE) in 100 days to like 70 ad the price I paid for them, buy then may not even do that. I think the best plan is to use the 16 Mhz and the resistor change far cheaper and effective. If I could get them to go to 28 Mhz would be lie having 7 for the price of 3 might be OK but I don't think I'll make it. As a friend of mine said "your pushing the limits of man and machine"

I would elevate you base heatsink a bit to get better air flow. maybe with pencils or something.

[mbbc]

...plus as USB hub controller...

What controller IC are you using?

[mjgraham]

My latest incarnation, decided to build a circuit board, plus as USB hub controller plus an aftermarket dc-dc , working the speed back up, I have found the problem that is causing my high speed issues seems to be ground loops, I guess it is to be expected with so many random parts in play. seems to be on the USB side I guess the long cable to the raspi that has a ground too. looks neat but doesn't work as well as I had hoped. I actually had all these parts already for many projects in the past. Have a single clock with a fanout buffer to the others. Plus all the work involved for such little gain really it changed it from ROI (just for the BE) in 100 days to like 70 ad the price I paid for them, buy then may not even do that. I think the best plan is to use the 16 Mhz and the resistor change far cheaper and effective. If I could get them to go to 28 Mhz would be lie having 7 for the price of 3 might be OK but I don't think I'll make it. As a friend of mine said "your pushing the limits of man and machine"

I would elevate you base heatsink a bit to get better air flow. maybe with pencils or something.

That is a good idea, I'll try that, amazingly the top heatsink is the one that is getting the hottest right now, I am having some troubles with the middle one getting more errors, I think my top mounting arangement is causing a high spot over the top one, I am going to go back and use thermal pads on the top, I basically used a small 3/8" square x1/8" piece of aluminum on top of all the BE100s to space it up tall enough to mount the top heatsink on and clear everything. If I put to much pressure on them they get detected as different devices or go away, I guess I am shorting some signals on the bottom. I am not sure why the top one got 40 PH/s must have been really fast

...plus as USB hub controller...

What controller IC are you using?

Some old TI TUSB2046B's just full speed only but that is enough

[Wed]

Thanks for this, i will try it!

[ElitePork]

Anyone has the power that the OCBE is drawing? I got 6-8watt on the wall measured for each OCBE which means each OCBE draws 1.2A to 1.6A... Which actually burnt my 4A adapter...

[ElitePork]

My little 8 puppies. 7 overclocked, 1 stock


They are so hungry that they ate up the adapter...




I got angry and decided to drown one to warn all others...

[bitminterlt]

Has anyone had an issue where the PADs for the oscillator came off of the PCB?

On two of the 20 units I did the 16Mhz/1.2K mods on, two out of the four pads for the oscillator came off with the SMD??.....


So two casualties out of the 20..... lol

Anyone had this happen to them, and has ANYONE been able to get them to work again with replacing the pads? I can upload pics shortly if anyone needs or wants to see it... This happen to one of the older units and one of the newest versions of the USB miners.

Thanks in advance!!

[ooxtcoo]

Has anyone had an issue where the PADs for the oscillator came off of the PCB?

On two of the 20 units I did the 16Mhz/1.2K mods on, two out of the four pads for the oscillator came off with the SMD??.....


So two casualties out of the 20..... lol

Anyone had this happen to them, and has ANYONE been able to get them to work again with replacing the pads? I can upload pics shortly if anyone needs or wants to see it... This happen to one of the older units and one of the newest versions of the USB miners.

Thanks in advance!!

from where are you?
i have a pcb, but need to solder new asic and new silabs chip.
if you near to me (austria) i could solder your silabs and you asic into my pcb and you have a working erupter again.

[mjgraham]

Has anyone had an issue where the PADs for the oscillator came off of the PCB?

On two of the 20 units I did the 16Mhz/1.2K mods on, two out of the four pads for the oscillator came off with the SMD??.....


So two casualties out of the 20..... lol

Anyone had this happen to them, and has ANYONE been able to get them to work again with replacing the pads? I can upload pics shortly if anyone needs or wants to see it... This happen to one of the older units and one of the newest versions of the USB miners.

Thanks in advance!!

well you can solder a small wire strand, just the oscillator pad or the whole trace to the BE100? you should see the mess I have with one. I pulled all the parts and stripped it , and mapped it out, trying to put it back together pulled a trace on the silabs chip and on the BE100 , had to make another board to hold the silabs , got the BE100 back on and none of it works

[ooxtcoo]

why is it not working?
is not detected by the computer?

i have repaired now so mutch demaged erupters...

if it is not detected and silabs is getting hot, then its:
a.) demaged silabs chip
b.) demaged asics

if you give the erupters to mutch current, the first think what burns is the silabs chip.
if you had luck, the asic is still working (if current wasnt to mutch).
if you have bad luck, silabs and asic is demaged.

if you change silabs chip with new one, and silabs is still getting hot, then asic have a short and is demaged.

if you get 000000 on detection in cgminer, then also asic is demaged.

if you have accepted shares but also many hw errors, it can have many reasons...
a.) bad asic (i have 2 asics, which are working but produze many hw errors)
b.) if you give not enugh ampere to the erupter (maybe to mutch erupters on 1 hub)
c.) some bad soldered resistor


if you get not detected erupter at all, check the usb connetor pins, sometimes they are not good soldered.


if you whant to have your erupter always detected with same comport number or if you have conflicts,
write your own serial number into the silabs chip and you can connect the erupter in any usb port and
you will get always the same port number again without need to install drivers again and again.


iam working now on a standalone asic miner with raspberry pi and 16x2 lcd build into a metall case and 5volt fans.
iam working now in the programming part, everything can be set directly with 2 buttons and the 16x2 hd44780 lcd.
you can set wlan, pool adress, pool username, pool password, backup pools and so on...
so you never need to connect the system to any other computer, everything can be set directly on the raspberry itself.

when iam finished with it, i release the source and the shematic to build it yourself.
its very simple to build, just a 16x2 (or any other) lcd, some wires and some resistors, 2 push buttons...



here is the video: http://jau.cc/lcd/cgminer_LCD_NEW.MOV
thanks to demonmaestro how uploaded it for me to youtube: http://youtu.be/mgocVRJrkNY


here is my old first video, with 16x1 lcd: http://jau.cc/lcd/cgminerLCD_ooxtcoo.MOV

[Zeek_W]

I screwed up with getting the wrong type of crystals. But my test erupter works fine now I replaced with stock xtal. Now I await a new shipment of xtals.

On the resistor front - damn I wish jaycar stocked them instore

Also I found with the xtal's - either flipping them and using a fine tip iron to add a little solder (or putting a little on the pads) then placing them on the board and using a heat gun to melt the solder works great.

EDIT:

An idea about getting enough power to a stick:

Anyone tried using those dual USB cables commonly found in older external 2.5in HDDs? Obviously your USB hub would be then out half their ports - but maybe a neat way to run it without screwing about with more indepth modding.

And to the person who posted the melted PSU - how random! I just had a dream last night where my pi's case melted like that!

[Trillium]

Why not just use the 5v rail of any ATX PSU. It's going to have better stability (regulated voltage output) than any $1 chinese generic power source that comes with a hub.

If you are worried about something shorting and the PSU dumping tens or hundreds of amps into the wires / BE circuits, add a inline fuse holder with some fast blow fuse(s). They are like $1 each from any electronics store.

[Scorpia]

I screwed up with getting the wrong type of crystals. But my test erupter works fine now I replaced with stock xtal. Now I await a new shipment of xtals.

On the resistor front - damn I wish jaycar stocked them instore

Also I found with the xtal's - either flipping them and using a fine tip iron to add a little solder (or putting a little on the pads) then placing them on the board and using a heat gun to melt the solder works great.

EDIT:

An idea about getting enough power to a stick:

Anyone tried using those dual USB cables commonly found in older external 2.5in HDDs? Obviously your USB hub would be then out half their ports - but maybe a neat way to run it without screwing about with more indepth modding.

And to the person who posted the melted PSU - how random! I just had a dream last night where my pi's case melted like that!

Jaycar? you must be aussie.

im in Melbourne and have spare resistors and a few spare crystals. but RSonline sells packs of 50 resistors for $2 delivered if you can find the right brand. I think they were listed as 4c each min 50 units.

delivery built into the price and they were here in 2 days. cant complain.

as for powering I find im just putting less per hub atm till I can get new power supplies. so instead of 5 I might get 4 etc.