Block Erupter USB - Overclocking/ hacking ?

[GilesBathgate]

So I am looking through the bfgminer driver: https://github.com/luke-jr/bfgminer/blob/bfgminer/driver-icarus.c

I can't see any evidence of error detection or correction on the protocol, no parity, no checksums, no resend.

The protocol just sends 4 bytes of nonce data. Ok maybe "the nonce *is* the checksum" since it will be checked by the host standard c sha256 routines. But all bfgminer does, in this case, is flag HW error and close/reopen the usb port. I would expect it might at least ask for the nonce again in case of transmission error.

[Bluestreak66]

So I am looking through the bfgminer driver: https://github.com/luke-jr/bfgminer/blob/bfgminer/driver-icarus.c

I can't see any evidence of error detection or correction on the protocol, no parity, no checksums, no resend.

The protocol just sends 4 bytes of nonce data. Ok maybe "the nonce *is* the checksum" since it will be checked by the host standard c sha256 routines. But all bfgminer does, in this case, is flag HW error and close/reopen the usb port. I would expect it might at least ask for the nonce again in case of transmission error.

As far as I know all the current mining hardware just passes along the nonce and then it is checked against sha256. None of the Asics on the market store the nonce for any amount of time as soon as it is generated it is passed along, if something happens in the process it is lost. When there are multiple asics communicating in a chain to a single micro if multiple nonces are generated at the same time they merge on the communication line and are lost. Correct me if I'm wrong but doesn't usb communication send parity check bits along with the data? It would seem that the silabs chip would handle some error checking on the usb layer but I may be wrong. I haven't dealt much with the software side of things when it comes to bitcoin mining. It seems like as long as error rates are less than 1% them nobody seems to bother reducing them further, I don't think I would bother either. However it would be good to know where these errors are coming from.

[DobZombie]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?

do I have to upgrade the power?

I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

[Bluestreak66]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?

do I have to upgrade the power?

I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

Here is a filtered list of oscillators on digikey of different freqencys http://tinyurl.com/m54mczl

Specs need to be: 3.3v, 5mm X 3.2mm, 4 lead SMD. The only ones that are pin compatible are the Tri-state versions with the OE pin. Check the data sheet to make sure it's pin compatible. The ones in the above link meet these specs. The one I am currently using is http://tinyurl.com/q98ua2j I also have a 18.43 and 20mhz version but have yet to test as I'm waiting on power supply parts. And yes power will need upgraded. Replacing just the crystal without any other changes put the board over the current limit for usb spec and at 1.05 volts it will not be anywhere close to stable. Read my previous post for voltages. Your going to need a hot air gun at the very least if you want to replace these with any success.

[haxtormoogle]

Bad news... I popped it

I had it working with a 20 mhz crystal for 5 minutes still was testing the hash rate... then it stopped responding. Now cgminer cant identify it, even with the stock resistors and crystal in place

I felt so bad, why I didn't post anything in a while.

On another note I figured out my clock gen circuit was putting out 5v not the 3.3 it should have been. Chances are this is what killed the chip not the speed.

still though I popped it

Got 3 more coming in the mail soon

I wonder if I could get a replacement chip cheep?

[SRoulette]

Bad news... I popped it

I had it working with a 20 mhz crystal for 5 minutes still was testing the hash rate... then it stopped responding. Now cgminer cant identify it, even with the stock resistors and crystal in place

I felt so bad, why I didn't post anything in a while.

On another note I figured out my clock gen circuit was putting out 5v not the 3.3 it should have been. Chances are this is what killed the chip not the speed.

still though I popped it

Got 3 more coming in the mail soon

I wonder if I could get a replacement chip cheep?

I am very impressed with your efforts so far
This thread is already way beyond what I could hope to discover by myself.

Paste a BTC address and I shall send you a small donation for your efforts.

[Bluestreak66]

Bad news... I popped it

I had it working with a 20 mhz crystal for 5 minutes still was testing the hash rate... then it stopped responding. Now cgminer cant identify it, even with the stock resistors and crystal in place

I felt so bad, why I didn't post anything in a while.

On another note I figured out my clock gen circuit was putting out 5v not the 3.3 it should have been. Chances are this is what killed the chip not the speed.

still though I popped it

Got 3 more coming in the mail soon

I wonder if I could get a replacement chip cheep?

What core voltage were you running it at? Ya I'd say the 5v clock signal probably done it in. I'm sure these will run stable at 20Mhz the only issue will be cooling. Do you have any estimates on temps or power consumption? Really sucks it cooked but thats just the price we pay sometimes.   Keep us updated!

[DobZombie]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?
do I have to upgrade the power?
I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

Here is a filtered list of oscillators on digikey of different freqencys http://tinyurl.com/m54mczl
Specs need to be: 3.3v, 5mm X 3.2mm, 4 lead SMD. The only ones that are pin compatible are the Tri-state versions with the OE pin. Check the data sheet to make sure it's pin compatible. The ones in the above link meet these specs. The one I am currently using is http://tinyurl.com/q98ua2j I also have a 18.43 and 20mhz version but have yet to test as I'm waiting on power supply parts. And yes power will need upgraded. Replacing just the crystal without any other changes put the board over the current limit for usb spec and at 1.05 volts it will not be anywhere close to stable. Read my previous post for voltages. Your going to need a hot air gun at the very least if you want to replace these with any success.

Rightio, so I'm just here to dumb this down a bit (he he he ) and make this a touch easier to understand for the layperson that wants fiddle with this sorta thing.

so first I should piss off the solder mask for better conductivity? like so...

Then I'd have to put down a very thin layer of an electrically insulating, but thermal conducting material.  This would prevent shorting on the now exposed PCB?
I could then add heatsinks and such to transfer more heat as needed?

Then I could replace the 12Mhz crystal oscillator with either...
a 14.31818MHz OSC which would produce approx 397Mh/s
or
a 16MHz OSC which would produce approx 444Mh/s
or
a 18.432MHz OSC which would produce approx 511Mh/s

So far so good?

Now I assume I'll have to replace a resistor or 2 to increase the power from 1.05v to 1.15v (or more?) and at what point (of OSC MHz increase) does the voltage have to be increased?

Now assuming I'd need a hot air rework station like this one to complete all this?

How does this sound so far?

AMIDOINITRITE?!?

[stellan0r]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?

do I have to upgrade the power?

I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

I would really like to try that, too

http://www.conrad.de/ce/de/Search.html?search=smd+quarz+oszillator+5+x+3.2+x+1+mm

just go straight for the top one, any other mods needed?  


(ok ok.. that one https://secure.reichelt.de/Oszillatoren-SMD-Keramikgehaeuse/XO53-16-00000/3/index.html?;ACTION=3;LA=446;ARTICLE=85031;GROUPID=4004;artnr=XO53+16%2C00000)

[GilesBathgate]

Correct me if I'm wrong but doesn't usb communication send parity check bits along with the data? It would seem that the silabs chip would handle some error checking on the usb layer but I may be wrong. I haven't dealt much with the software side of things when it comes to bitcoin mining. It seems like as long as error rates are less than 1% them nobody seems to bother reducing them further, I don't think I would bother either. However it would be good to know where these errors are coming from.

Its not really a USB connection, its a rs232 serial connection. rs232 can support parity, but this is just for error detection. What I would do, if I had my hands on the firmware would be to modify it to send the 4 bytes of nonce, 3 times in a row (so total of 12 bytes). This would serve as error checking (the 3 nonce values didn't match) and error correction (take the two nonce's that do match). There are more complex schemes for checking and correction codes, that potentially use less bytes, but for the sake of 12 bytes this is about as simple as you can get.

[ooxtcoo]

Bad news... I popped it

I had it working with a 20 mhz crystal for 5 minutes still was testing the hash rate... then it stopped responding. Now cgminer cant identify it, even with the stock resistors and crystal in place

I felt so bad, why I didn't post anything in a while.

On another note I figured out my clock gen circuit was putting out 5v not the 3.3 it should have been. Chances are this is what killed the chip not the speed.

still though I popped it

Got 3 more coming in the mail soon

I wonder if I could get a replacement chip cheep?

yes, i think for your work we all should spend you another usb asic miner for your (our) tests you make!

if it is not detected, then i think it have nothing todo with the asic itself...
cause it should be detected even without the asic soldered on board.
the silabs usb to com emulator should handle it... i only know use ftdi in my flasher box, but i think its works the same way.

[Bluestreak66]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?

do I have to upgrade the power?

I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

I would really like to try that, too

http://www.conrad.de/ce/de/Search.html?search=smd+quarz+oszillator+5+x+3.2+x+1+mm

just go straight for the top one, any other mods needed?  


(ok ok.. that one https://secure.reichelt.de/Oszillatoren-SMD-Keramikgehaeuse/XO53-16-00000/3/index.html?;ACTION=3;LA=446;ARTICLE=85031;GROUPID=4004;artnr=XO53+16%2C00000)

That oscillator should work fine, it's pin compatible, voltage and size is correct.

Correct me if I'm wrong but doesn't usb communication send parity check bits along with the data? It would seem that the silabs chip would handle some error checking on the usb layer but I may be wrong. I haven't dealt much with the software side of things when it comes to bitcoin mining. It seems like as long as error rates are less than 1% them nobody seems to bother reducing them further, I don't think I would bother either. However it would be good to know where these errors are coming from.

Its not really a USB connection, its a rs232 serial connection. rs232 can support parity, but this is just for error detection. What I would do, if I had my hands on the firmware would be to modify it to send the 4 bytes of nonce, 3 times in a row (so total of 12 bytes). This would serve as error checking (the 3 nonce values didn't match) and error correction (take the two nonce's that do match). There are more complex schemes for checking and correction codes, that potentially use less bytes, but for the sake of 12 bytes this is about as simple as you can get.

Are you talking about the communication from the micro to the silabs chip? That is rs232. If your talking from the silabs chip to the pc that is an emulated rs232 connection. Error checking is done per the usb protocol. Each USB transaction has a Token Packet, Data Packet and Status Packet. With those three things it can transfer the data, check the data for errors and correct errors that happened during that transaction. Error correction does not work 100% of the time. I can pull the firmware off the micro if your interested in looking at it.

[ooxtcoo]

you dont think that fuses are blown in the atinny?
i dont think you can dump the firmware.

[Bluestreak66]

you dont think that fuses are blown in the atinny?
i dont think you can dump the firmware.

I'm not sure, has anyone tried? I don't really have any need to do so anyway, so I'm not going to bother unless it would be of use to somebody. I just figured I offer to try.

[ooxtcoo]

if we cant mod firmware for more power (cause its static), then its only good for repair..
i dont tryed to dump the firmware, but mostly fuses are set to prevent clones.

[Bluestreak66]

if we cant mod firmware for more power (cause its static), then its only good for repair..
i dont tryed to dump the firmware, but mostly fuses are set to prevent clones.

Other than looking at the the communication protocol in the firmware I really don't see a need to have it or modify it. There would really be no reason to lock down the micro as there are no BE100 asics out there to build a clone with. I would almost bet the fuses are left intact. I can't get my programmer to play nice with windows 8 atm anyway so I'm not going to bother. I can confirm that the pinouts are correct as was posted, as my programmer does recognise the attiny but I'm having driver issues with the pc. 

[haxtormoogle]

I took everything I learned so far and make a post on my website: http://wtfmoogle.com/?p=3334 How to overclock your Block Erupter and Possibly Destroy it in the process.

[Bluestreak66]

So can somebody link me to the 14mhz and the 16mhz crystal oscillator that can be used?  can I just buy any old 4 pin crystal?
do I have to upgrade the power?
I'm a little thrown off by some of the specifics of the thread, and I want to get in start soldering (I'm awesome with a soldering iron )

Here is a filtered list of oscillators on digikey of different freqencys http://tinyurl.com/m54mczl
Specs need to be: 3.3v, 5mm X 3.2mm, 4 lead SMD. The only ones that are pin compatible are the Tri-state versions with the OE pin. Check the data sheet to make sure it's pin compatible. The ones in the above link meet these specs. The one I am currently using is http://tinyurl.com/q98ua2j I also have a 18.43 and 20mhz version but have yet to test as I'm waiting on power supply parts. And yes power will need upgraded. Replacing just the crystal without any other changes put the board over the current limit for usb spec and at 1.05 volts it will not be anywhere close to stable. Read my previous post for voltages. Your going to need a hot air gun at the very least if you want to replace these with any success.

Rightio, so I'm just here to dumb this down a bit (he he he ) and make this a touch easier to understand for the layperson that wants fiddle with this sorta thing.

so first I should piss off the solder mask for better conductivity? like so...

Then I'd have to put down a very thin layer of an electrically insulating, but thermal conducting material.  This would prevent shorting on the now exposed PCB?
I could then add heatsinks and such to transfer more heat as needed?

Then I could replace the 12Mhz crystal oscillator with either...
a 14.31818MHz OSC which would produce approx 397Mh/s
or
a 16MHz OSC which would produce approx 444Mh/s
or
a 18.432MHz OSC which would produce approx 511Mh/s

So far so good?

Now I assume I'll have to replace a resistor or 2 to increase the power from 1.05v to 1.15v (or more?) and at what point (of OSC MHz increase) does the voltage have to be increased?

Now assuming I'd need a hot air rework station like this one to complete all this?

How does this sound so far?

AMIDOINITRITE?!?

Sorry I missed your post, but yea thats that how it would go. As far as Voltages/frequency I plan to post a chart up of what works @ what frequency, of course your mileage will vary. Not all chips are created equal some will require more voltage and some will require less but if we can get a good sample going I think we can create a base line to go off of. I ordered some oscillators and some other parts to make it easier to tweak voltage than by replacing resistors. I'm still waiting for one of my orders.

I took everything I learned so far and make a post on my website: http://wtfmoogle.com/?p=3334  How to overclock your Block Erupter and Possibly Destroy it in the process.

Thats a good summary I don't think it was the 20mhz that killed it rather the 5v clock source. I got a variety of oscillators up to 24mhz so we shall find out soon enough. 

If anyone knows of any adjustable bench power supplies or cheap oscilloscopes for sale let me know. I'm transitioning jobs coming up here in a few days and won't have access to the equipment I do now so I'd like to buy some for hobby use. PM me if you know of any for sale.

[joeventura]

I asked the same question on your website,

Can you just swap the crystal to 14 or 16 and nothing else and get increased results?

[Trillium]

I asked the same question on your website,

Can you just swap the crystal to 14 or 16 and nothing else and get increased results?

A good rule in life is: No, you can't get something for nothing. There is always a catch.

In this case you risk destroying your miner quickly, or possibly reducing its life. As the number of people who've tried changing the oscillators can be counted on one hand, we have no idea how long these modified units can run for. YMMV.