Do we have to get the surface mount crystals? Would a can style one stuck top of the inductor be OK, or does it have to be close to the board?
I was also wondering if all 4 pads are different, or if they are doubled up? I thought crystals only needed 2 contacts.
I have also seen voltage and capacitance mentioned - do these matter?
Topic: Block Erupter USB - Overclocking/ hacking ? - page 30. (Read 168765 times)
@Trongersoll:
Hot Air gun (a cheap with regulated temperature) works well... it's a bigger Problem to get a replacement part available for "everyone". I can get some from my Company, how about you?
@bobsmith652:
that's why you get those weird numbers. Those resistors have usually 1% tolerance and fit those printed values.
believe me, i earn my living with such stuff.
Hot Air gun (a cheap with regulated temperature) works well... it's a bigger Problem to get a replacement part available for "everyone". I can get some from my Company, how about you?
@bobsmith652:
that's why you get those weird numbers. Those resistors have usually 1% tolerance and fit those printed values.
believe me, i earn my living with such stuff.
is it safe to assume that someone at home with a soldering iron doesn't have much of chance of successfully swapping out that crystal?
Yes, I measured them in situ. The numbers on the top plug into the formula nicely.
Also, crystal measurement (scale in mm on my kid's plastic ruler)
Probably closer to 4.8mm x 2.5mm
Also, crystal measurement (scale in mm on my kid's plastic ruler)
Probably closer to 4.8mm x 2.5mm
You did not measure them while soldered in, did you? 
R1 & 2 are clearly marked 7500 and 2401, making it "750 with Zero additional Zeros" and "240 with one additional Zero", at least on that pic of V3 i found somewhere.
R1 & 2 are clearly marked 7500 and 2401, making it "750 with Zero additional Zeros" and "240 with one additional Zero", at least on that pic of V3 i found somewhere.
I don't know how you came up to 1.16?
Datasheet tells Pin 4 is Feedback and should be around .8 V.
Resistors R1 (750 Ohm) and R2 (2k4) seem to be the Voltage divider, R2 being the lower part.
Calculation is .8V/2k4 * (2k4+750), makes perfect 1.05V for me.
You'll need to replace the 750 Ohms with 1k2 to get 1.2V, 1k giving 1.13V, 1k1 makes 1.16.
Crystal seems to be in 12SMX size, but i can't check, i don't have any Sticks yet.
Questions:
- Which size is the Crystal?
- Which frequency can be safely obtained (sufficient cooling necessary, but we want max hashpower)
- Which voltage is needed for this?
Also, you don't Need two USB ports to power 1A. Buy a cheap hub, they usually have no fuses, just connecting all 5V from the Outputs, the power supply and the incoming cable(!), often making your Computer not shut down properly. All you need is a beefy psu.
On the other Hand: if you can resolder the stick to hash faster, you also can resolder the hub to supply more.
I plan to Bridge all Outputs and supply with a switching reg capable of 8A on a 10-porter Hub, that should bring enough power for overclocked erupters and maybe for backpowering a RPi.
Datasheet tells Pin 4 is Feedback and should be around .8 V.
Resistors R1 (750 Ohm) and R2 (2k4) seem to be the Voltage divider, R2 being the lower part.
Calculation is .8V/2k4 * (2k4+750), makes perfect 1.05V for me.
You'll need to replace the 750 Ohms with 1k2 to get 1.2V, 1k giving 1.13V, 1k1 makes 1.16.
Crystal seems to be in 12SMX size, but i can't check, i don't have any Sticks yet.
Questions:
- Which size is the Crystal?
- Which frequency can be safely obtained (sufficient cooling necessary, but we want max hashpower)
- Which voltage is needed for this?
Also, you don't Need two USB ports to power 1A. Buy a cheap hub, they usually have no fuses, just connecting all 5V from the Outputs, the power supply and the incoming cable(!), often making your Computer not shut down properly. All you need is a beefy psu.
On the other Hand: if you can resolder the stick to hash faster, you also can resolder the hub to supply more.
I plan to Bridge all Outputs and supply with a switching reg capable of 8A on a 10-porter Hub, that should bring enough power for overclocked erupters and maybe for backpowering a RPi.
Resistor measurements from my earlier post:
R1 = 603R
R2 = 1058R
R3 = 7.45K
R4 = 3.36K
Measured using 2 different multimeters.
The end 2 pads on the end of the board have 0v and 5V on them. The others all have a mixture of 3.13V and 1.25V on them. You could probably use the end 2 to power the board.
Edit: I just measured another board, and got
R1 = 605
R2 = 980
R3 = 7.45K
R4 = 7.45K
Dunno why R1 and R2 are not what they say on the top. Looking at the numbers, I am more confident about choosing new values though.
The 12MHz crystal is 5mm x 2.5mm.
I don't know how you came up to 1.16?
Datasheet tells Pin 4 is Feedback and should be around .8 V.
Resistors R1 (750 Ohm) and R2 (2k4) seem to be the Voltage divider, R2 being the lower part.
Calculation is .8V/2k4 * (2k4+750), makes perfect 1.05V for me.
You'll need to replace the 750 Ohms with 1k2 to get 1.2V, 1k giving 1.13V, 1k1 makes 1.16.
Crystal seems to be in 12SMX size, but i can't check, i don't have any Sticks yet.
Questions:
- Which size is the Crystal?
- Which frequency can be safely obtained (sufficient cooling necessary, but we want max hashpower)
- Which voltage is needed for this?
Also, you don't Need two USB ports to power 1A. Buy a cheap hub, they usually have no fuses, just connecting all 5V from the Outputs, the power supply and the incoming cable(!), often making your Computer not shut down properly. All you need is a beefy psu.
On the other Hand: if you can resolder the stick to hash faster, you also can resolder the hub to supply more.
I plan to Bridge all Outputs and supply with a switching reg capable of 8A on a 10-porter Hub, that should bring enough power for overclocked erupters and maybe for backpowering a RPi.
Datasheet tells Pin 4 is Feedback and should be around .8 V.
Resistors R1 (750 Ohm) and R2 (2k4) seem to be the Voltage divider, R2 being the lower part.
Calculation is .8V/2k4 * (2k4+750), makes perfect 1.05V for me.
You'll need to replace the 750 Ohms with 1k2 to get 1.2V, 1k giving 1.13V, 1k1 makes 1.16.
Crystal seems to be in 12SMX size, but i can't check, i don't have any Sticks yet.
Questions:
- Which size is the Crystal?
- Which frequency can be safely obtained (sufficient cooling necessary, but we want max hashpower)
- Which voltage is needed for this?
Also, you don't Need two USB ports to power 1A. Buy a cheap hub, they usually have no fuses, just connecting all 5V from the Outputs, the power supply and the incoming cable(!), often making your Computer not shut down properly. All you need is a beefy psu.
On the other Hand: if you can resolder the stick to hash faster, you also can resolder the hub to supply more.
I plan to Bridge all Outputs and supply with a switching reg capable of 8A on a 10-porter Hub, that should bring enough power for overclocked erupters and maybe for backpowering a RPi.
Just looking @ those picture and reading whats on the linked site Id say whats going on here is they have built these down to usb spec power consumption. That 12mhz crystal drives the ASIC and the internal PLL multiplies that by 28 (12*28=336. If you look at their site it says they work stable at 392mhz (which 14*28=?). I would even bet that one could swap a 16mhz crystal in to get 448mhz. Of course voltage would need adjusted. R1 and R2 create a voltage divider and compare that against an internal reference regulate the voltage. So just adjust the voltage to ~1.2 maybe a little higher via resistors and you should be able to have a nice piece of equipment. What I would do is bolt 2 or 3 to a cpu cooler spaced so that they plug into a usb hub cut the 5v traces to the usb ports in the hub and solder a 5volt supply directly to the boards and run the cpu cooler off the 5 volts as well. That would be a nice tiny cheap 1GHash/s rig. Is the other side of the pcb flat under the heatsink?
In A previous post, I calculated the voltage from the resistor measurements should be 1.16V. Measuring across one of the resistors gave me 1.04V, which is close enough to friedcat's quoted 1.05V to give me confidence that I have the right 2 test points.
So another thing I am not too confident about is what resistor values to use, as the 1.16V and 1.04V don't add up for me. There appears to be a voltage drop, but I wouldn't know what values to use to compensate for this.
The underside of the board, once you unscrew the heat spreader, is a plain aluminium flat surface.
Seems it depends if you're confident enough to desolder the surface mount resistors and replace them for higher voltage as well as the crystal.
I'm not going to try. Last time i tried, the buggers stuck to my soldering iron. I'm guessing the SMT crystal will be a bugger to desolder as well as it has pads underneath. A good candidate for ripping up tracks right there.
Just looking @ those picture and reading whats on the linked site Id say whats going on here is they have built these down to usb spec power consumption. That 12mhz crystal drives the ASIC and the internal PLL multiplies that by 28 (12*28=336. If you look at their site it says they work stable at 392mhz (which 14*28=?). I would even bet that one could swap a 16mhz crystal in to get 448mhz. Of course voltage would need adjusted. R1 and R2 create a voltage divider and compare that against an internal reference regulate the voltage. So just adjust the voltage to ~1.2 maybe a little higher via resistors and you should be able to have a nice piece of equipment. What I would do is bolt 2 or 3 to a cpu cooler spaced so that they plug into a usb hub cut the 5v traces to the usb ports in the hub and solder a 5volt supply directly to the boards and run the cpu cooler off the 5 volts as well. That would be a nice tiny cheap 1GHash/s rig. Is the other side of the pcb flat under the heatsink?
Seems it depends if you're confident enough to desolder the surface mount resistors and replace them for higher voltage as well as the crystal.
I'm not going to try. Last time i tried, the buggers stuck to my soldering iron. I'm guessing the SMT crystal will be a bugger to desolder as well as it has pads underneath. A good candidate for ripping up tracks right there.
I think there is one clock for the controller and one for the hashing unit.
the 12 Mhz is used by the hashing unit and it will be internally multiplied to 256 Mhz or something like that.
the 12 Mhz is used by the hashing unit and it will be internally multiplied to 256 Mhz or something like that.
Seems it depends if you're confident enough to desolder the surface mount resistors and replace them for higher voltage as well as the crystal.
Using 2 USB slots is a good idea, means you don't have to go messing around inside the USB hub too, gives you up to an amp which is plenty, just make sure you get a bigger cooler - might be up to 4w.
2.5w*1.15*1.15*(14.7/12)=4.05w
(15% more voltage, therefore 15% more current, times (new clockspeed divided by old clockspeed [because of extra number of switches, which is the time in which the chip produces heat])=new output wattage.
Might want to add a ramsink to the VRM too once you've modded it.
Using 2 USB slots is a good idea, means you don't have to go messing around inside the USB hub too, gives you up to an amp which is plenty, just make sure you get a bigger cooler - might be up to 4w.
2.5w*1.15*1.15*(14.7/12)=4.05w
(15% more voltage, therefore 15% more current, times (new clockspeed divided by old clockspeed [because of extra number of switches, which is the time in which the chip produces heat])=new output wattage.
Might want to add a ramsink to the VRM too once you've modded it.
When asked about overclocking the blades, friedcat had this to say:
When asked about software overclocking of the usb sticks:
Seems like there is room to play, especially if you take an additional 100mA from a second usb port after swapping out the 12Mhz oscillator and upgrading the heatsink.
Quote
The oscillators are fixed on the hashing board (one 12MHz, one 14.318MHz). Low switches to the 12MHz one while High switches to the 14.318MHz one. If you need more speed you will need to buy a 3.3V 5mmx7mm oscillator that is higher than 14.318, unsolder the current one, and solder the new onto the board. We tried 14.7xx and 15 before. On 1.25-1.26V they begin to work, but producing a lot of heat.
When asked about software overclocking of the usb sticks:
Quote
No possible because the oscillator for the USB is a fixed frequency one. The voltage is also fixed at the lowest possible for the current frequency to match the 500mA spec of USB ports.
Seems like there is room to play, especially if you take an additional 100mA from a second usb port after swapping out the 12Mhz oscillator and upgrading the heatsink.
I like the look of these too http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=251195651419 the clip could touch the ASIC with the main flat on the underside heat spreader.
A bit slow on delivery, but they do fit if you splay the tab slightly, as seen in the pic below:
They haven't got a large enough recess to reach sideways, so they clip on the end and touch the 2 chips. The heat spreader seems to be a lot cooler than the others, although the sink doesn't feel warm. I have always had a fan on them, and this seems to work well. Obviously you need to space your devices out more with these on.
Has anyone had any thoughts on the overclocking? Is the 11 or the 12MHz the one to replace?
Nice work, mine have just been ordered and I will be hackling away and mining alt coins shortly
I already have a nice collection of clock crystals from previous projects so I hope to get testing quickly.
There is an 11 and a 12 MHz surface mount crystal on the Rev3. The 11MHz is the 3rd device from the left at the top in that photo that has the flash whitewashing the markings out. Not sure how that makes 333MHz...
just a shot in the dark, but 11*12*2.56 = 338 perhaps that is somehow responsible for the 333MHZ speeds? (or am i just 'making math work'?)
Where did the 2.56 come from?
I was wondering that too, normally to multiply a clock signal, a PLL is used to multiply the input by a certain integer.
We need someone more versed in hardware design to have a close look, I wouldn't be comfortable definately recommending replacing something and breaking your erupter, although you could get a 13 or 14 mhz crystal and try both, not sure if this would be safe/cause issues though.
I'd suppose if you replaced the USB one by accident it just wouldn't be recognised by the computer till you switched it back out. Try PMing friedcat.
There is an 11 and a 12 MHz surface mount crystal on the Rev3. The 11MHz is the 3rd device from the left at the top in that photo that has the flash whitewashing the markings out. Not sure how that makes 333MHz...
just a shot in the dark, but 11*12*2.56 = 338 perhaps that is somehow responsible for the 333MHZ speeds? (or am i just 'making math work'?)
Where did the 2.56 come from?
There is an 11 and a 12 MHz surface mount crystal on the Rev3. The 11MHz is the 3rd device from the left at the top in that photo that has the flash whitewashing the markings out. Not sure how that makes 333MHz...
just a shot in the dark, but 11*12*2.56 = 338 perhaps that is somehow responsible for the 333MHZ speeds? (or am i just 'making math work'?)
I like the look of these too http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=251195651419 the clip could touch the ASIC with the main flat on the underside heat spreader.
A bit slow on delivery, but they do fit if you splay the tab slightly, as seen in the pic below:
They haven't got a large enough recess to reach sideways, so they clip on the end and touch the 2 chips. The heat spreader seems to be a lot cooler than the others, although the sink doesn't feel warm. I have always had a fan on them, and this seems to work well. Obviously you need to space your devices out more with these on.
Has anyone had any thoughts on the overclocking? Is the 11 or the 12MHz the one to replace?
To work out the exact Vout to the ASIC either measure voltage out across pin 3 and the 'far' side of R1 whilst running (riskier), or measure the resistance values of R1+R2 whilst off and check against that datasheet. (pinouts and supporting circuitry p12/13, voltage programming table p7)
R1 = 603R
R2 = 1058R
I make that 1.25V or 2.2V, depending on which way around the 2 would be. They aren't K ohm though.
R3 = 7.45K
R4 = 3.36K
These give 1.16V with the formula, but I can't tell where the vias go as the other side of the circuit board is an aluminium plate, so I am not sure which of these 4 are the ones. R3 and R4 are the only ones in the K ohm range out of the 4 nearby though. I can see R1 is the one on pin 4, so it must be one of the 2 in the divider circuit.
Measuring the voltage from pin 3 to the opposite side of R1 (where the via is) gives 1.04V.
IDK WTF is going on, but then I am a software engineer who has to drive hardware, not design or debug it.
Jump to: