Topic: Block Erupter USB - Overclocking/ hacking ? - page 4. (Read 168713 times)

Pretty crazy, I did try a peltier for a while, I found it everything gets to dependent on to many things. I had three that ran at 26 Mhz but one failed after a week and it seemed to kill the BE100 chip, voltage was pretty high ~1.3v or so I think it is over +7A on the low voltage. I had to put a (big) heatsink  on top and bottom, although with the peltier you could transfer more heat but as I was corrected a few messages back, these things use 6~8x more power to run at 2x the speed and it does cost a lot more plus all the added complexity but it was a fun experiment at any rate. All I used was air cooling for the setup would be interesting if you could actually lower the temp like that if it would help , one thing I noticed was the area that the heat is coming out of is really small, there is a lot of wasted cooling going into the board , especially if you remove the on board regulator. Glad to see someone is still tinkering around with them mine quit this past weekend when the power supply went out in the signal generator.

Yea, this is my kinda (universal) power supply rail I use at work to test different devices. I just wanted to really test out the controller ... and since it does not have PID functionality, it kind of sucks and will only lead to headache if try to use it with multiple peltiers. That being said, I am going to get a PID unit and see how many BE's I can mount to a steel plate. It would be interesting to see if I can get a stable 700MH+ out of them since I have about 20 of them. Waiting on some 30A 5VDC Power supplies at the moment.

Weren't you doing some BE experiments previously?

Oh and the peltiers on eBay take forever for delivery if coming from Hong Kong or wherever. Worth it to pay an extra couple bucks and get them from Amazon Prime (Free 2-day Shipping) 

Holy crap, that is a lot of solid state relays and other gear just to control one pelitier.

There are peltier controllers on eBay for quite cheap ($<20), or you could wire up the internal relay (1 amp max @ mains voltage usually) of any standard model Eurotherm controller to a ATX PSU & peltier. These controllers go cheap on eBay as well (check part / model number against spec sheet to check what installed options you are buying though).

Just thought I'd mention... Asicminer is soon releasing a 49 port hub, maybe that'll make things easier for some of you and be an attractive product. You can see a picture of it here http://bitcointalk.org/index.php?topic=99497.14440

Started bench testing a Peltier cooler purchased off of Amazon and a temperature controller. Should have went with a PID loop control - maybe will change if I work up to multiple Peltier coolers. Anyway, I did do some testing with one of my OC'd BE's running the same at ~450MH. I can get the temp down to around 15C without any condensation problems. The only problem is that with this temp controller, there is a 1 minute compressor delay built in, so when I hit 15C, it cuts off, waits one minute, and then applies power. When it is off, the temperature is around 30C on the BE right before the Peltier kicks in again. When I upgrade my one BE to over 20Mhz Xtal and bypass the voltage regulator, I am sure I will need all the cooling I can get! Will keep things posted.

What did you use for the Xtal and R1?

Well I got 2 done.

I think I need more power.  the hubs I received could not put out enough power to run more than 2 anyway.  Need to get creative

My BE run at ~400 mega, all I did was solder the usb power supply directly to a 5v line in a computer PSU. It outputs 5.1 or 5.2 volts.

Though, when I was originally testing, the power dropped to 4.5 volts and thry still got ~400.

Here is a 30A 5VDC Power Supply pretty cheap. You can also use a computer power supply too. Digikey, 237-1306-ND

Love it 

My first overclocked BE, using scrounge parts - that's a 16.2MHz oscillator I found on some 20+ year old not-sure-what-it-used-to-be in a miscellaneous box.




My current little garden of OC'd BEs - this collectively converges on just shy of 6GH. It's powered off the 5V rail of an old 350W ATX supply. I didn't notice any overall increase in reliability from adding the chip heatsinks - HW errors didn't decrease. Got a few more coming in but I probably won't overclock them until I get another hub - or start removing the chip heatsinks.




I beefed up the hub rails and capacitors but was still having power delivery issues, so I wired all the hot pins straight to the 5V input. I don't trust barrel connectors for high current, so I'd also already hardwired the hub boards to power supply connectors.

Here is some good basic knowledge to know that might help in your tinkering:

"A lot of this information was collected from mjgraham, bluestreak66, and trillium on this forum. So thanks "

• 1) ASIC Clock Speed = (Xtal) * 28
• 2) Stock ASIC Clock Speed= (12Mhz*28=336Mhz)
• 3) Stock Xtal is 3.3V@12Mhz, 5mmX3.2mm, 4Lead, Tri-State/Open Type
• 4) ASIC Chip MAX Voltage is 1.8VDC
• 5) Regulator (AZ1021) Output = 1VDC @ 2.5A, (3A MAX @ 100% LOAD)
• 6) V=0.8(1+(RA/R2))     Stock R2 Value is 2.4kOhms
• 2) Stock ASIC Clock Speed= (12Mhz*28=336Mhz)

Here is also some additional information that I have collected:

• 1) Adding a 10uF capacitor across the input 5VDC may help clean up power/stability.
• 2) The higher the frequency of the clock, the more voltage required to stabilize them.
• 3) Current draw increases with frequency, NOT voltage.
• 4) For high clocks, it is necessary to replace the regulator with an external high amp buck converter that is adjustable from (1.05V-x).
• 5) Add the following to cgminer: "cgminer (--icarus-timing short=90)"

Here mjgraham's table for reference:

If the inline panel meter is in any way accurate, they pull in the neighborhood (probably higher) of 1A off the 5V line at 16.384MHz
I took the generic 10-port hubs and ran new power wires from the 5V input to the 5V terminal on the USB jack in a star configuration, so there wouldn't be excessive current loading (and associated heat and voltage drop) on the traces, works pretty well.

Pushing very far past that operating frequency will require more power than the onboard regulator can source reliably, so either replacing the drive chip and inductor with a 5A (vs 3A) is necessary, or wiring around it with an external high-current regulator. I forget the name but someone a few pages back did a full range power test with these guys, up to above 24MHz. Good data.

I do not have a BE here at work.  can someone measure the distance between contacts on the back.  I need to know what pitch connector to get.


Thanks guys.

Yes, assume each BE will draw 2A with Overclock. Wire as many as you can in parallel to your power source, but not to exceed the Output Current Max of the PSU. You can cut open a USB cable and wire the data lines and GND to the BE as well. Just make sure to leave off the 5VDC wire so you do not backfeed into your hub/computer. You should quadruple check your connections before powering up, and then measure your voltages before connecting the USB to the hub/PC.

You can just take apart any USB hub and wire directly to the BE's. Leave off the 5VDC wire from the USB Hub and Power all the BE from an external supply that can provide enough amps.

With those giant heatsinks, why don't you overclock them to something higher than factory clock? Maybe dial it up past 18Mhz and see if you can get past 600MH. You may need to supply external 5VDC 2A Power Source for each unit (bypassing the USB connector).

Got a baker's dozen of these guys running 16.384MHz (~460MH/s) off 1.35V - probably a little higher than it needs to be, but the first ones done have been running solid for about a month. Few more coming in the mail, but they won't get overclocked until I have more hubspace. Not really for profit so much as for entertainment - had a lot of fun learning the hardware and doing the mods. I probably got in too late/too small for BTC to really be profitable, but what I have invested has resulted in hundres of hours of entertainment. The first BE I tried to hack, I hosed the regulator and killed something. The next one was ugly; didn't have any surface-mount parts so wired up 1/2W resistors for the feedback adjust and my oscillator was an ancient four-pin thru-hole spider'd onto the pads. Stuck up like an inch off the board, fell off a few times, ripped off some pads and traces, thought I'd killed that BE at least twice but it's still running.


Got one of the new blades (the ones without adjustable VRMs) running stable off the same speed oscillator; pushing out about 14.8GH. Should have parts kits available end of the week if anyone's interested, and a full writeup is pending a bit more playing around. For my own playing, gonna be putting together a fully adjustable clock source and see just how high it'll go. Probably put up pictures of the whole setup tomorrow sometime.

Hilarious!

I have just a basic hot air tool. I set it to around 300C to do the work. You can find some youtube videos to show you some smd techniques. I suggest getting some flux in a syringe and some isopropyl alcohol. I clean the are with the ISO, then apply some flux around the areas i want to desolder. Then, I start to heat with my air gun using tweezers to put tension on the component. The solder melts and I can remove the component easily. Now, I apply a bit more flux and heat up the solder pads so they become nice and fresh. Let the PCB cool for a bit, clean the area with ISO, apply flux and slowly start to heat the pcb. Using tweezers, position your component and then apply the hot air. Once a little solder melts, then apply pressure from the top of the component and heat around the pins. This way, the component will solder and seat flush and secure with the PCB. It's really not as complicated as it sounds. Takes maybe 5-10 minutes per board and that is with me being OCD on the soldering technique.