Pages:
Author

Topic: Block Erupter USB - Overclocking/ hacking ? - page 21. (Read 168765 times)

hero member
Activity: 615
Merit: 502
Thanks I was using the external 15A adjustable voltage supply with the higher frequency crystals so current should not be a problem. 

I had ordered the crystals I'm currently using off eBay and the specs were not too clear.  Now they don't seem to work the same as yours I've gone and ordered the ones off DigiKey.  I hope I can get some higher speeds from these as I've bought a lot of cooling equipment.

For now I've upgraded 9 of my 10 erupters to run at ~360MHash with the eBay 16Mhz crystals while I'm waiting on the the DigiKey crystals.

member
Activity: 70
Merit: 10
I've got my 15A external power supply now and it's working with a 16Mhz crystal ok, I'm getting avg. 360MHash/s with this installed, running at 1.35V - is this a normal speed for a 16Mhz crystal?  Little or no hardware errors with this config.

If I put a 20Mhz or 24Mhz crystal in, the light doesn't go off and the BE isn't recognised by the BFGminer software.  Does that mean it's the wrong type of crystal or any suggestions as to how I could make the higher speeds work, even for a short time?


I got a 447Mhs with a 16mhz sometimes it would drop into the low 430's and this was at 1.21v so 360Mhs seems low. You wont be able to use a oscillator higher than 18.43 with the onboard core voltage supply, it goes into overcurrent protection mode. What happened for me was that the Block Erupter was recognized by the pc but once hashing started it would trigger the overcurrent then timeout. 18.43mhz would do the same after a few minutes so I have no idea what voltage it would be stable @ above 16Mhz.
hero member
Activity: 615
Merit: 502
I've got my 15A external power supply now and it's working with a 16Mhz crystal ok, I'm getting avg. 360MHash/s with this installed, running at 1.35V - is this a normal speed for a 16Mhz crystal?  Little or no hardware errors with this config.

If I put a 20Mhz or 24Mhz crystal in, the light doesn't go off and the BE isn't recognised by the BFGminer software.  Does that mean it's the wrong type of crystal or any suggestions as to how I could make the higher speeds work, even for a short time?
newbie
Activity: 24
Merit: 0
Oh man, this thread is fantastic. Great work guys.

I have some experience in soldering, but very little in smt and those crystals look to be very tightly packed in there. Is it better to use a hot air solution to solder them in?

Also, i'm now wondering about power modding hubs so that they can handle several of the overclocked miners, plugged in as they were intended. Still thinking about cooling and such, but i cant help feeling i want to build something water cooled. Perhaps using something like these:

http://tinyurl.com/mvytnq4
http://tinyurl.com/m3g9m9n

Anyway, keep up the great work fellas. The thought of building an overclocked usb miner rig will now dominate my every thought.

hero member
Activity: 896
Merit: 532
Former curator of The Bitcoin Museum
Hope this helps guys Smiley

unfortunately there is no "Like" Button on this page.

Your positive comment is better than a like button Grin
hero member
Activity: 574
Merit: 500
Medical Translations for Bitcoins
Hope this helps guys Smiley

unfortunately there is no "Like" Button on this page.
hero member
Activity: 896
Merit: 532
Former curator of The Bitcoin Museum
Well, after I asked a couple of times what resistors I'm going to need to attach to R1 to change the voltage, I got off my lazy bum and did a bit of research Smiley

After reading Moogle's Blog

He had this picture showing the Voltage Regulator and the resistors...


I looked through the Voltage Regulator Datasheet that moogle linked.

In the datasheet, there was an equation for Voltage Output

V = 0.8 ( 1 + ( R1 / R2 )

Now I had NO IDEA how to read the codes on the resistors, so I learnt from From this page

Basically, you read the first 3 digits as the number you need, and then the 4th digit is a reference to the multiplier. So...

4700 = 470 × 1 = 470Ω
2001 = 200 × 10 = 2,000Ω or 2kΩ
1002 = 100 × 100 = 10,000Ω or 10kΩ
7503 = 750 × 1000 = 750,000Ω or 750kΩ
3504 = 350 × 10000 = 3,500,000Ω or 3.5MΩ

So R2 is 2401 which means 2.4kΩ (or 2400Ω)
and R1 is 7500 which means 750Ω

Therefore to calculate voltage...

V = 0.8 ( 1 + ( 750Ω / 2400Ω )
V = 1.05

So to figure out what resistor to replace R1 with to change the voltage, here's a table...

|VOLTAGE||R1 Resistor to Use||Resistance in Ohms|
1.05v
legendary
Activity: 2212
Merit: 1001
oh wow this cant be right. what is my block erupter doing?
Holly cow!
This can't be correct

I was running a stability test and iit just started mining like crazy!
did cgminer just go stupid or something?

OMG what did you do it's sentient now! Run! RUUUN for your life!!
 Cool

http://www.youtube.com/watch?v=bPmVhyHBRAM
hero member
Activity: 574
Merit: 500
Medical Translations for Bitcoins
oh wow this cant be right. what is my block erupter doing?
Holly cow!
This can't be correct

I was running a stability test and iit just started mining like crazy!
did cgminer just go stupid or something?

OMG what did you do it's sentient now! Run! RUUUN for your life!!
 Cool
hero member
Activity: 910
Merit: 1002
Its a bug
newbie
Activity: 14
Merit: 0
oh wow this cant be right. what is my block erupter doing?
Holly cow!
This can't be correct
http://wtfmoogle.com/wp-content/uploads/2013/08/hollycow.png
I was running a stability test and iit just started mining like crazy!
did cgminer just go stupid or something?
hero member
Activity: 615
Merit: 502
Thanks for the voltage range updates Bluestreak66, my crystals arrived today. I've got 16MHz, 20MHz and 24Mhz here now.  My external power supplies aren't here yet but am going to try 16MHz tonight with the built in supply. 
member
Activity: 70
Merit: 10
Couple of updates:
Block Erupter is Stable @ 447/1.2v~1.25 with relatively low heat dissipation compared to the previous ~1.4v
20mhz exceeds current limit and voltage regulator goes into overcurrent mode when mining begins.
18.43mhz crystal maxes out onboard regulator and does not produce stable results at any voltage over long periods of time.

I do believe the voltage regulator is damaged at this point I cannot get it to output a stable voltage even back @ 447/1.2v, it just goes into overcurrent mode. So I do believe thats it for tonight, I need to come up with a solution for a higher current regulator. At Least now I think we have a good voltage range @ 16mhz.



member
Activity: 70
Merit: 10
Got 3 new block erupters to play with Smiley



Can you guess what one is overclocked Cheesy

Using a 16.386mhz crystal and replaced r1 with a 1.2kohm resistor. Put a large heatsink on and everything looks stable


dare I try 18.43mhz? Smiley

Awesome good to see your back in the game! 18.43? NO LETS TRY 24mhz!  Shocked

Whats your measured core voltage?
newbie
Activity: 14
Merit: 0
Got 3 new block erupters to play with Smiley

http://wtfmoogle.com/wp-content/uploads/2013/08/fastbe.png

Can you guess what one is overclocked Cheesy

Using a 16.386mhz crystal and replaced r1 with a 1.2kohm resistor. Put a large heatsink on and everything looks stable


dare I try 18.43mhz? Smiley
member
Activity: 70
Merit: 10
Possibly it might be better to use an SPI controlled VCO

There are some really 'cheap' solutions available. That way you can have a bit of software that increases the clock rate until the thing starts producing too many errors, then you just gradually 'slide' the VCO back down a few tens of  KHZ

Take a look at silicon labs, they have a full range of products that can be picked up on ebay or group buys fairly cheaply.

What you will see is that the number of errors will increase as the frequency increases, so with a VCO you can 'trade off', also as the temp is a function of the speed you can slide the VCO Downwards if you start getting too hot, whilst still mining.

Only 'fly' i can see is If there is some sort of clock syncronicity between the USB circuit (UART/USB bridge) and the ASIC
My approach here is that I would like to get a stable voltage range for oscillators that are easily attainable from well known trusted sources (such as digikey, mouser, element14, ect) and post a list of these with resistor values. That way a oscillator and resistors can be purchase and install with some certainty that it will work. Having a dialed in frequency at which is not obtainable in a oscillator package is not very useful. Now if we had these in bare chip form and were making custom boards we would be having a different conversation. I am more interested in dialing in the voltage at a given frequency as I think that will be more useful. The only road blocks here are I see are heat dissipation (which I think are mostly in check at this point) and current delivery capability.

Far better to go for an internally regulated part, trying to get a very accurate external voltage is not easy and usually requires additional voltage regulation chips, plus there is no indication of the conditions a user is running under. Just because a test setup works with a particular clock setup is not really a solid indication that other setups will work the same, or indeed be non destructive.

As long as everyone realizes that you would have no feedback control, that is to say if the component values you choose trashes a certain number of ASICS, there in nothing you can do to control it, other than unplug the power from the device.(unless you have some sort of clock gating circuit)

I'm not particularly pro or con what you are trying to do (hell I've toasted more post 500$ chips than most), but it is important that you take into account variation of the devices.


As regards the inductor, if you say it is cool... then it is cool, but it is a little more complicated than just setting a reference.



Sorry I'm still not following you. I'm using the voltage regulator thats on the board here is the data sheet: http://www.aosmd.com/res/data_sheets/AOZ1021AI.pdf It samples the output voltage to keep it steady, it's closed loop operation, the output is contiouosly monitored . The resistor divider is used only to offset the voltage to the comparator and has no effect on the system other than the output voltage look at page 7 of the pdf. It's a synchronous buck design so the output stage consist of 2 switches, an inductor and smoothing cap, nothing else.
sr. member
Activity: 399
Merit: 250
Possibly it might be better to use an SPI controlled VCO

There are some really 'cheap' solutions available. That way you can have a bit of software that increases the clock rate until the thing starts producing too many errors, then you just gradually 'slide' the VCO back down a few tens of  KHZ

Take a look at silicon labs, they have a full range of products that can be picked up on ebay or group buys fairly cheaply.

What you will see is that the number of errors will increase as the frequency increases, so with a VCO you can 'trade off', also as the temp is a function of the speed you can slide the VCO Downwards if you start getting too hot, whilst still mining.

Only 'fly' i can see is If there is some sort of clock syncronicity between the USB circuit (UART/USB bridge) and the ASIC
My approach here is that I would like to get a stable voltage range for oscillators that are easily attainable from well known trusted sources (such as digikey, mouser, element14, ect) and post a list of these with resistor values. That way a oscillator and resistors can be purchase and install with some certainty that it will work. Having a dialed in frequency at which is not obtainable in a oscillator package is not very useful. Now if we had these in bare chip form and were making custom boards we would be having a different conversation. I am more interested in dialing in the voltage at a given frequency as I think that will be more useful. The only road blocks here are I see are heat dissipation (which I think are mostly in check at this point) and current delivery capability.

Far better to go for an internally regulated part, trying to get a very accurate external voltage is not easy and usually requires additional voltage regulation chips, plus there is no indication of the conditions a user is running under. Just because a test setup works with a particular clock setup is not really a solid indication that other setups will work the same, or indeed be non destructive.

As long as everyone realizes that you would have no feedback control, that is to say if the component values you choose trashes a certain number of ASICS, there in nothing you can do to control it, other than unplug the power from the device.(unless you have some sort of clock gating circuit)

I'm not particularly pro or con what you are trying to do (hell I've toasted more post 500$ chips than most), but it is important that you take into account variation of the devices.


As regards the inductor, if you say it is cool... then it is cool, but it is a little more complicated than just setting a reference.

newbie
Activity: 6
Merit: 0
So with my current setup I have a 20Mhz Crystal installed and using a trim pot I can get it to run but I cant get it to go ant faster the the 330ish that it runs at normally with the original crystal . I spliced in 5v directly from PC power supply and I am going in through the USB cable with the power.  at the moment all I can say is its running cool attached to an old CPU heat sink so no heat issues yet. i think I need more power but if I turn the trim pot down it errors out and stops responding.
full member
Activity: 151
Merit: 100
It's the HW value from CGMiner.
should be max 1% of the A + R value, i think.
Some HW's are ok, but should get not too much.


I got

Code:
A: 146 R: 0 HW: 19

thats not to bad in my opinion, but i'm not expert

You need to do the calculation with diff 1 shares. If your shares are not diff 1 then you need to convert them to diff 1. If your pool is set to 4, 8 ,128 diff for example, you need to do the math in that case.

I have 20 Block Erupters running stable (for well over a week) on 2 - 10 port USB Hubs and cgminer is reporting the following hardware errors for the last 12 Hours with a diff of 4.

Code:
A: 17465 R: 34 HW: 979

What is the calculation from diff 4 to diff 1?  Or better yet, would you consider the reported HW within the normal range?
newbie
Activity: 6
Merit: 0
resistor R1 is the one you want to change but you have to change it with a resistor that has a value that will match with the crystal you are using to run the asic chip at the desired speed. Bluestreak66 is the one who knows some of these working combos so far.

I will be reciving some stuff to test with today and will post my results. Even if I push it to hard and blow up my BE.
Pages:
Jump to:
© 2020, Bitcointalksearch.org